 Well, thanks very much for the nice introduction. Thanks for having me here in Singapore. Actually, you told the complete story of what I'm about to tell you guys, actually. My TV here in the end is centered around my topic today, how open source, open data, and the cloud are reshaping finance education and the financial industry. I think I'm old enough to have experienced this breakthrough that open source and the other technologies that we will talk about today have brought to the industry. Back in the days when I wrote my PhD thesis quite a while ago, almost 20 years, we had a completely different landscape. So I'm still really excited about the things that are going on in our industry and what we see today. And although I'm having here kind of a rather general title in a sense, of course, I will focus on Python for two reasons. One is that this is a Python conference. The other one is that it's exactly what we do. But of course, I will touch upon other open source technologies as well. Let me start by quickly saying a couple of words about myself in addition to the nice introduction before. I'm indeed founder of the Python Quants company. We are active globally, although we are small, operating out of Germany. We have kind of reached, I wouldn't say global scale, but global reach at least with conferences that we do in London, in New York, and things that we are planning now for Asia as well. So it's kind of nice to see that our topic Python for Finance has become such a global phenomenon. I'm really excited about that. When I started using Python for Finance, many people said, well, this is not the right language to do finance actually, because it's interpreted. It was back then maybe given what we see today, it was rather nascent, although it was Python where we started using it was already around for 20 years or something. So it's not something new. I still meet people that say, well, there's kind of a new thing. I never heard of it. That's kind of amazing, because a million people around the world are using it, and we'll talk about this later on as well. Again, we are a small company, but we're doing Python for Finance, at least with a global reach. And if you're writing books, and I'm in constant touch with people from every continent that you can imagine from Australia, South America, the US, and Europe, and obviously in Asia as well. What we mainly did over the last couple of years is doing client work. This is one of the projects that I can mention, because there's no NDA or no things attached. It was more or less a marketing initiative for our German derivatives exchange, UX, where we built open-source-based tutorials for V-stocks-related derivatives and also variants futures. And open-source in this case means, indeed, that the majority of the stuff was presented and implemented in Python. And this was kind of a nice strategic decision in the sense that I said, well, we want to give everybody who is interested in our products a chance to use it for free, and not to have something where you need to buy a license, where you have kind of vendor lock-in. We don't know how the future will evolve, so let's at least decide for an open-source technology and within this realm of open-source technologies. I mean, I could have been another part here in this regard, but they decided for Python, I was really happy for that they started presenting kind of on a technological level, concepts related to their financial products based on Python, and we were able to support them. On the other hand, what we, what I'm doing is speaking, giving trainings, traveling around the world. This is one picture where I gave a tutorial in Berlin last year at the Big Data Conference. So this was not actually Python focused, not a Python conference, but Python is playing more and more a role in these big data fields, fintech fields, and so forth as well. So I'm really happy to see this as well. If you're interested and if you are around on Saturday afternoon, I'm giving a similar tutorial where we make use of open-source technologies to attack typical problems in financial data science, like building a web service from scratch, building a couple of nice, simple web applications that it can build with Python rather efficiently. But if you're interested, I've also prepared kind of a more fintech-oriented example of using Python for a Bitcoin related algorithms, actually. Our vision, when I started using Python, it was more that I saw it as a tool, a tool in a sense that, yeah, I thought and in the end, I was proven right that it would make finance much easier in the sense that Python as a language with its syntax is so close to math and everybody was a little bit of formal education and a little bit of programming experience and understands, I can read financial equations, is really able to understand what Python is about when they see it with regard to financial algorithms. So when I lectured at the university, most of the students think they haven't had any kind of proper programming experience, but nevertheless, they were able to catch their ideas and the implementation approaches based on Python. Of course, we didn't do that many advanced things, but seeing the financial equations, the algorithms, knowing a little bit of financial backgrounds and then seeing the implementation in Python proved kind of a good combination. And today, we've evolved to a state where I see us working on a revolution with regard to financial education, research and practice through Python. So now, it has turned around and we try to spread the word. We do more and more trainings, I get involved in university, things like building or designing a new master degree which centers around open source and finance and so forth. So we would say, when we go through the material of the day that is now more kind of spreading the word and bringing it through the world than before when we started out where we thought, well, this is a nice tool, let's do it for what we want to do. So it was more kind of the egocentric thing, now we are more spreading the word and bringing more and more things around the globe, actually. Overview and agenda, I have a couple of topics. I wanna start with a couple of thoughts about the data science and finance fields as of today. I will speak about a little bit of the revolution that open data science in the cloud bring along. Then I covered the importance of the Python ecosystem. Based on a couple of statistics, I will answer the question or try to answer the question, why Python for finance at all? And elaborating a little bit more on my thoughts from before in a sense that Python is kind of well suited to be a companion for any kind of formal discipline, actually. Finally, and this is the largest part here, it's how to learn and do Python for finance. I mean, the one thing is to see that something is kind of nice and easy to use. The other thing is how can I scale that? How can I bring this to a university class? How can I bring this to bigger companies, to bigger teams and how can I make it easy that such a technology gets applied? Also in the long run, I mean, strategic decisions at the academic level, university level as well as in the corporate level are typically not made for a time horizon of six months. So if a company, for example, decides to use Python as a strategic platform, then there must be something behind it in the sense that the decision is made for five years plus, more or less for 10 years in general, at least. So this is this major part. And finally I will conclude and a little bit restack things that I've presented before and show kind of what we've been working on and I will present the complete package with regard to derivatives analytics with Python. Data science and finance today, I mean, if you want to become a data scientist today and you read kind of the requirements or the job profiles, it seems that you need to have kind of three, four, five master's degrees and two, three PhDs. So you're getting more and more inquiries where companies are looking for PhDs with 10 years of experience knowing these technologies and more and more and more. So it seems to be kind of a few but only Superman can survive or people like Albert Einstein or other very famous intelligent people. But in the end, I mean, we live in the world where technology obviously plays a role. On the one hand, always saying, especially to my son that math is very important. Unfortunately, he has fun doing math and languages. I mean, I will later on come to that what relates Python to English when it comes to languages and programming languages respectively. But here it's kind of indeed a field that is grown rapidly and the requirements for people who want to work in this field have grown at the same pace, actually. And if you can't believe the statistics then to be data scientist seems to be the sexiest job in the world. I think there are a couple of things that make it indeed attractive if it's sexiest job in the world, I don't know. I'd rather be a famous DJ. I think this is more sexy being a data scientist and doing data science as a hobby maybe. What drives this evolution? Of course, we all know that I just decided to present something that I find kind of a nice picture. The source is down there. I will point you to the slides later on. Of course, data is growing. It's getting faster and faster with the growing. So obviously, exponential growth. Actually, and people want to make something out of that. I mean, the one thing is to have data. The other thing is to make something out of that. Even 50 years ago, you could have had 500 books but if you haven't read one of these, they don't bring anything. So how to read the books, how to make sense of the data that you have. This is kind of indeed a thing and I think we are still here in the beginning in the sense that many, many companies are very, that they rush to new technologies, they have big budgets, they hire people but it's still not kind of entrenched with traditional business models. I'm not speaking about kind of the new startups that built from scratch things on top of the data that they collect. It's more or less kind of the corporate world which still dominates. Although the impression, if you follow the press, might be a little bit different that startups are kind of reigning the world. This is still not the case. So we will see there many, many more things in the future and this field where we get nothing that is a hype but more or less kind of entrenched with all the other disciplines that we see kind of finance reporting, controlling and so forth. On the other hand, and this is particularly true for our finance field or our, I mean, what we are working in, it's kind of this tremendous shift. So when I, back in the day, that my PhD thesis, almost all that was available and this was even hard to get your hands on, it was end of day data, a sense that the complete theory, when you have a look at the famous theories like mean variance portfolio theory or the capital asset pricing model, any kind of other thing, was more or less based on end of day data observations that have been made there. But this has shifted tremendously in the sense that we now have data available in real time and at high frequency. I mean, this is one of the reasons why we have big data in the first place is that we now measure things in a much more granular fashion. But on the other hand, in order to survive at the markers, and I get up from many, many people who are interested in kind of algorithmic, automatic trading, high frequency trading and so forth. So this seems to be kind of a fascinating field, but this brings along, obviously, a couple of other requirements. Even just thinking of the technologies that you need in order to cope with kind of real time data or think of the internet of things, but here I will stick with my financial examples, it needs completely different technologies, protocols and ways of coping with the data. So this is another requirement, not only did the volume itself with the frequency that we see here, the velocity as they typically have the second V. And in particular for our finance field, this is true when people need to cope and need to deal with masses of data these days in our field, as well as with the speed that the data is generated. And here I'm speaking more or less when you see such pictures of structured data, which is kind of the simple case. We just have time series data, which is well structured and can be well managed. If you then now think that news is also dropping in kind of a high frequency manner, but it's unstructured data, then the requirements might get even a little bit higher. On the other hand, at least speaking of our financial industry, and if you believe it or not, if you're not active in the financial industry, Excel is still kind of the most popular tool there. And many, many people are trying to even manage risk management and lead banks more or less based on Excel. So also for the foreseeable future, this would kind of be one of the major things. And if you now have the angle of view of a risk manager, of a compliance officer, whatsoever, and have a little bit of understanding what technology is all about, then you see kind of the risk that we have there. So think of version control for spreadsheets. There are a couple of approaches, of course, but how do you control what the trader is doing, tweaking his Excel sheet or whatsoever? And we all know about the scandals where they were kind of one billion loss, 1.5 billion loss at a French bank and they just discovered it after months that 1.5 billion were lost, you know. It's kind of, yeah, because people have influence on the technology and there's no version control and you cannot really have a control procedure in place when you're using these technologies. On the other hand, traditional IT processes are inappropriate. I used to work in the financial industry in Germany with a focus on the insurance industry. And if you think of a life insurance, for example, what they typically have is kind of a time horizon of 10 to 30, sometimes 40, 45 years. That's all very, very long-term. My experience working in this industry for almost 10 years was kind of that the processes, decision-making processes and so forth have kind of a similar horizon. In the sense that, well, this is all for the long-term and if you build something, this needs to last for like 10, 20 years, but we all know that the cycles, the product cycles and technology cycles are getting shorter and shorter and shorter. And in a certain sense, and there's a reason why I like to cite Henry Ford here is if I had asked people what they wanted, they would have said faster horses. And this is exactly what I've seen in this industry and other industries as well, that they are stuck in kind of a status quo and just try to marginally improve certain aspects of what they are doing and not kind of thinking, well, what are others doing? What is here? And I think, especially in this industry, we will see a little bit more of a disruption. And if you see the cycles, I mean, insurance is now coming a little bit. But in other industries, we have seen it as well. And why do I am picturing here the iPhone, having myself an iPhone is kind of, it's a typical example that nobody ever asked for an iPhone ever came up with it. And now we think this is kind of the standard, you know? We're looking for things that they will come up with in the future as well and sense that, yeah, traditions might not be what we shall use. And more or less, think outside the box, don't ask the market, just think for yourself and see what you can do. It's also the same, getting so many questions regarding what books shall I read in order to do this and that, usually I say, well, master the basics, math, technology, and so forth, and then think for yourself and see what you can do. I mean, you can read 100 books, but if the wisdom would indeed be in these books, in these finance books, then we would have many, many more millionaires and billionaires who have made kind of the money by reading the books and applying the theories. This is kind of common knowledge and obviously doesn't bring that much in the end. The rise of open data science, now getting a little bit more deeper to kind of singular trends that we observe in the market that are kind of heavily influencing, in particular our industry and not the industry, but also from my point of view, and this is what I find really exciting these days, is kind of the financial education. Coming myself out of this, I was mentioning, writing a PhD, thesis in math finance, and having had the hard time to get my hands on data, actually. First of all, of course, there's GitHub, there's open source, and yesterday I really enjoyed the keynote of yesterday with regard to open source, what open source can bring, and what kind of mass movement this has become. GitHub more or less replacing in the development world kind of the CVs for developers. You just point people to your GitHub account and say, have a look at what I did. It's not that you write about writing code, you can show what you have written, and you can really illustrate what you are capable of. So this is more like a musician where I say, well, these are my kind of 10 albums or 20 records that I did, and there's no talking about it, you just have a look. And on the other hand, if you wanna learn something, they're not as many secrets as they have been. So 10, 15 years ago, especially in the finance field, you haven't had a chance to have a look at kind of any financial code. This was simply not around. If you're interested how companies are doing option pricing or implementing other financial algorithms, there was no way to look for it. There have been a couple of books where you saw it kind of on a rather abstract academic level, but not to have a look at how others, professionals are doing the work. And now you have the chance. You can look at, to stay in our domain, if you're interested in pandas, data science, or what server, you can have a look at it because it's simply open source. So the one thing is kind of illustrate and to print things out, the other one is kind of that you have your library now open where you can study what others have been doing. Of course, they're kind of different qualities and different scopes. And so, but this is kind of the beauty of this as well, that you have all these options and these vast sources to learn and to have a look. And also this platform to present yourself actually. On the other hand, and this is for me, similarly amazing, especially for our finance field, which was kind of over the years influenced by the fact that financial data has always been costly to get. So when I wrote my seminar thesis before doing my master's studies in Germany, it's actually a diploma, I had to pay my professor 20 Deutschmark to get a very tiny little data set. So 20 Deutschmarks for a student back then, it was kind of, it was money, it was a tiny data set. I got it on a disk and went back home and started analyzing this little data. Today, if you're studying finance or any other kind of field, you have platforms like Kwandl. Kwandl is kind of really popular now in our field in the financial industry. They are unifying more than 100,000 open data sources on the platform, providing the data via restful APIs, chasing, you get a chasing base, CSV, whatsoever. You have a nice Python wrapper, you import Kwandl and you can retrieve data from these 100,000 sources. And this is kind of nice. And the other example, some might be a little bit more popular, retrieving stock price data from Yahoo or from Google and so forth. So these days working in the field, it's indeed kind of quite easy to get your hands on the data and it's kind of fascinating. You have open source, you can use these fascinating tools and you have the data and you just buy a machine and you sit in your office writing your thesis or whatsoever and you have access to a vast amount of resources that haven't been available 10, 15 years ago. On the other hand, if you wanna speak to the experts, now come to the people. There are many, many meetups. I mean, I arrived here on Tuesday morning and once the night I went to PyData Singapore meetups, it was really nice to see the communities all around the world. And they asked me to give a brief talk and it's kind of this open communities, to not have kind of this setup which I learned to know as a professional in the beginning as well. We have to go to conferences if at all, pay a couple of thousands of euros or US dollars or wherever you have been located just to get access and then there have been a couple of speakers sitting there and couldn't really approach them. And now we have these open communities, we have these conferences that you can meetups typically for free and technology conferences, my experience at relatively low prices compared to the professional ones. So you have all these communities where you can exchange your ideas, where you can meet with people from other companies that might do similar things but you would never speak to because they are not your colleagues. And 10, 15 years ago, same thing, even with meetups seven, eight years ago, you could more or less only speak to the people you knew, which meant your colleagues or a couple of others that you might have met at university or whatsoever but now you simply go and have many people in my London meetup, for example, I say, well, I'm the only one who is doing Python for finance at my bank or in my department or whatsoever and really enjoying it here, meeting people that do the same thing. Before I would have had a chance and now I can simply go at night, don't have to take a day off for a conference and I can speak to other people and listen to talks where I learn so to say for free and maybe I even get a beer and a snack afterwards, kind of really nice. Getting back to the technology, also open source when you think of a browser and the browser being the operating system and Trubida Notebook, who have used using Trubida Notebook? Not everybody, I'm a little bit surprised. If you are not using it, please start using it this afternoon. If you come to my tutorial, or the Delaney's tutorial, Contorpion Tomorrow, I think gets about financial analysis with Trubida Notebook. Actually, I don't have that many examples. Tomorrow's more about programming in my tutorial but you must start using it. I live in this thing. You see, this is my presentation that I've written in Trubida Notebook. My last book or the next that will appear, I finished it, has been written completely in Trubida Notebook. So these are things that you can do these days that are amazing with Trubida Notebook so you should go and have a look if you are not using it for Trubida Notebook. Even if you are not the heavy Python user, if you're using another language like R or whatever, there are kernels available for Trubida where you can make use of other technologies apart from Python as well. But there are many, many other things. I mean, I indeed live in the browser. I could lose this thing today and set up something tomorrow. It takes me half an hour because I'm more or less only using the browser, Gmail, Dropbox, integration, Cloud. We come to that in a second. And this is kind of a recent days when you lost your machine and you haven't had a backup from yesterday, everything was gone. But now you have everything in the cloud and accessible typically while the browser and all the technologies like Dropbox. But have a look at Trubida. This is, if you don't believe me, but believe me that, this is something you must do. Cloud storage, I mentioned it before. It's kind of fascinating what cloud storage can do. I mean, we are not, I mean, how Dropbox works these days or they're working on the next generation is that of course you have redundant data storage in a sense that you have it typically locally and then it gets kind of replicated to the cloud and you can replicate it to other machines. But still for me, kind of it's kind of nice because it's our platform. I will speak about this in a minute or a couple of minutes. Having stuff stored here in my machine, working on that and this gets replicated to the cloud. Before I started using Dropbox, it was a heavy Git user also for my private stuff, not for coding. So I version controlled, if you like, it's not a real work at all because it's binary. I used Git as a backup system where I replicated and cloned my stuff. But now it's much more easy because you have the capacity of one terabyte of storage and simply save my stuff and when I work on the platform and no matter what infrastructure, someplace else, I just replicate this to our platform as well, which means I can for example have a local editor open and I change something in the code file and this gets on the server kind of in real time synced, so to say. Next generations will be even a little bit more convenient because they're now working on kind of a hybrid version. But the message here is more elastic. Cloud storage has been or got traction first in the private area but now it's getting more and more a corporate story as well. And we see this also revolutionizing how companies manage their IT infrastructure. Unlimited and affordable compute power. I mean, most of the business models and startup stories that we see today are only possible because there is kind of the unlimited and affordable compute power in the form of a cloud storage. Digital lotion is something that we use. You can start here with five US dollar per month for the smallest instance, having one core, 512 megabytes of RAM and 20 gigabytes of storage. This is something you can start with. I mean, you can do your development work there and if you need more capacity, then you can scale it. So it's in principle unlimited like in principle, electricity is also unlimited when you get it out of the cloud. And it's also kind of nice that you only pay for what you use and not have to also back in the day to be good at the company, you bought your big thing and either your lease, a term was like three to five years or whatsoever in here, just say, no, I need it for this hour or for 10 hours or we are doing your training, therefore I need the big machine and you just rent it for as long as you wanna rent it and this is what you pay for in the end. Kind of really nice and makes also for us, for example, we work for client and makes it really easy to set up things, easily scale and if you now think of EC2 and most of the traffic now is all over the world is due to companies like Netflix, I think Netflix is the company that generates most of the traffic in the world right now and they host their stuff on Amazon EC2 cloud. So this is how the world works. Also in the financial field, we would see more and more in this direction, not maybe that Deutsche Bank or any other big banks would put their data on digital ocean or Amazon, but they will use the technology for the buildup of internal clouds and make this also more elastic and can marginalize kind of the units that can be used by other departments. Now this was a little bit of the bigger picture as I see it. Now let us come a little bit more down to Python and what role Python plays in the bigger picture. In education and I'm really happy about to see this and this is already, I wouldn't say old, but older. You'll see it from July, 2014. And back then, almost two years ago, you see that already back then, most of the top 39 US computer science departments use Python to introduce programming. Second Java, I think maybe in 2012, Java would have been the leading language here. Matlab, C, C++, still scheme, scratch kind of designed for younger people, for teens, for even maybe smaller, like my son, nine years old, they can start using scratch, but I'm really happy that in education, Python is taking place. And although in finance, obviously, this is for the computer scientists, for those who will influence computer science in the long term in our future. On the other hand, when it gets to data science, what we see is that not only Python plays kind of a dominant role here together with R, it's kind of the rise of the open source technologies. When data science and statistics was dominated back in the day by technologies like SAS or others, in this regard, Matlab, we see SPSS, obviously, also a domain-specific language and tool. And now we see that open source languages dominate and open source technologies. Languages, the one thing, technologies and ecosystems are the other way. Say, well, this is not kind of like a standard language like C++, which is defined, but there's much, much more. There's so many more things around that. And usually when I speak of Python, I rather avoid speaking of programming language only. For me it's kind of a strategic technology platform and it's not only kind of a language, although the language itself has its beauty and I will come to that in the next part. Julia also, this is on the most right side from you. This is kind of really fast growing and has similar features and similar appeal from my point of view to Python. I've had a couple of speakers at my meetups as well about Julia and this has built in high performance, if you like. Many, many things you can achieve with Python as well, but here is kind of the language which is a little bit newer, obviously, but they have tried from the beginning to overcome this problem, but Python originally had kind of being an interpreted language and being rather slow in certain areas of application when you're not using kind of the performance technologies that are available these days. On the other hand, having a little bit more of a general look, you see here search trends on Google for tutorials for respective languages and there most of the languages are kind of declining in popularity, but some of them that I mentioned already before like R and Python that are here in the middle column, they are kind of increasing in popularity in the sense that people are looking more and more to get started and to learn these open source languages. And this is kind of, yeah, because a couple of elements play a role here. The one thing is that more and more employers are looking for people with a skill. The other side is that you can do amazing things with Python and R together or alone and that it's so easy to get started. So if you say, well, I wanna do something in Python or in R, just go to the website, download it and in principle, a couple of minutes later you have everything that you need to get started. There's nothing that you need to buy or a trial license or even to register or something. It's simply downloaded, start your tutorial and a couple of weeks later you might be already an expert. So happy to see that this is growing. On the other hand, of course, said it before, this is driven by the big companies and I watched the keynote by Kido van Rossem at the Python US, I think two weeks, three weeks ago, the day after it was recorded. And he mentioned, I wasn't aware of that, that the first implementation of the Google search and rank algorithm was implemented in Python. Obviously today this is not the case anymore but this illustrates, from my point of view, very well what Python can achieve. You get very efficient, very productive in a sense that once you have an idea, it's easily and quickly expressed in Python and tested. With other languages, we have the typical implement, compile, re-implement cycle. This might take a little bit longer but with Python and especially in finance, it's quite easy to implement ideas and you have to get quick results which is also really important because in the end, many people are speaking about high performance when it gets to something which is optimized and implemented and now you wanna scale it. But typically, the bottleneck for the majority of application and scenarios is kind of the developer, the quant analyst or this data scientist's time in the sense that if these guys need to spend weeks and weeks to come up with the first solution, this is kind of wasted time. Later on, it does make sense if you need six weeks order to implement something and then you make it high performance for a week of application, this doesn't make sense. If you have in Python have it implemented in five days and then the execution takes a week, you're much better off in the total time. So I think this is kind of the beauty but I'm not saying that Python is solving every kind of problem but of course a couple of them and Bank of America, they have even decided to build their core risk management system which is called Quartz based on Python. So analytics and pricing still is in C++ and as we know, Python can easily interface with C++ libraries but all the logistics around that, the Quartz system itself, which plugs then in the analytics implemented in C++ is built on Python and there are hundreds of people that even if not thousands of Bank of America that program Python on a daily basis and as far as I know, this is the biggest financial system in production based on Python, more than 14 million lines of code that Bank of America has deployed for their risk management system actually. So we see not only important Python in the world is also heavily used and Bank of America is just one example. So I'm here just confined to one I could mention, JP Morgan, many, many of the big hedge funds like AQR to Sigma, who are heavy Python users but I think the biggest example here speaks for itself. But why Python for finance at all? What brings Python with it? That makes it so beautiful, both for financial education and financial practice. I have two examples, the one is financial algorithm example. A little bit of Python code now here. I guess the details are not that important but most of you might be well-worsed in the technologies I'm presenting here. What I'm doing just to keep the code concise I'm indeed doing kind of a from pile up import stars. This is very rare, I think first time since five years this example that I'm doing it just to show how concise the code is in the end here. The example is the 101 of option pricing, the Blex-Cold's model, 1973, Geometric Pound Emotion. You see the sarcastic differential equation here and you see a difference equation which just doesn't really kind of do discretization over time, it's just like that you say the end value of my stock price index level or whatever I want to model by this model as given in the fashion as written down here. If you are not, if you don't know this model this is kind of not a big issue. The only thing that is noteworthy here is that we want to simulate by Monte Carlo simulation end of period values for again, think of an index or a stock price or of a currency exchange rate for a currency pair. This is what we want to do. And here is already the implementation so it's actually more text and more equations, two lines of equations and a couple of lines of text here and the Python implementation itself is, oh, you don't see the left hand side. So this is a little better. You see what is happening here. The first line, the very first line is just to time what is executed here. The second line is the Python code that starts. We just define a couple of variables, parameters and a mathematical financial sense. The next line of code set equals standard normal then we have 10 million as the number because we want to do 10 million simulations of the end of period value. No, think of doing this with axle, problem maybe. A single line of code to draw 10 million random numbers and in the last line we now apply our difference equation to calculate the end of period values given our parameterization and the standard normally distributed pseudo random numbers. And we are set after three lines of code and this whole thing takes on my local machine here. This is 2013 MacBook Air, still works, not maybe the fastest but it just takes 1.33 seconds. Maybe someone coming from a hedge fund or Big Bang would say, well, this is too slow but again, think of financial education and that everybody of us could easily replicate this after one minute of typing the code down. So kind of nice and something I say quite often is you have an equation, you translate the equation to Python and it's typically a one to one relationship but only when it gets to the syntax but with a number of lines of code. On the other hand, when we do financial data science we will see this typically when you have a kind of an analytical question and you translate this to Python and pandas this typically also translates into a single line of code. That's a kind of convenient and really, really effective and productive. Now a little bit of a check, so calculating the mean value that we would expect kind of a sanity check here would come up with a number of about 105.1 given our parameters for the model and this is kind of the analytical value I'm calculating here and this is the mean value. So having 10 million numbers calculated, I have it available as numpy and the array object, a vector here in a mathematical sense and just calling the mean method on that gives me the average. So I don't need to loop over everything, need to add something up and afterwards take the average. I simply have mean and don't need to take care of any kind of looping. Of course the looping takes place under the hood using numpy but this is rather quick because it's implemented in C. So we have the speed of C and the compactness and the pythonic way of doing things here on the Python level. So let me get back to the regular resolution. Plotting, another thing. So picture says more than a thousand words, we all know this expression dating back to 1911 already and you see here in principle, a single line of code would suffice but I'm doing a little bit of customization here. This line in the middle here is just getting the plot with the frequency distribution, the histogram of my data. 10 million numbers, give me the histogram. It's only a couple of characters to get the plot done. In the last line I'm just adding the mean value that we can check where the mean lies here in the histogram but you see even if you are not a computer scientist nor a programming expert, you're a beginning finance student, you should be able after a couple of hours of instruction, maybe a couple of minutes only to generate such plots and to implement such an analysis. Last but not least, getting to the next step, simulation typically is not done for the sake of the simulation, it is done to do something and for us this means more often than not, pricing of options. And here again kind of a nice illustration of how concise and efficient it is to write Python code. Here I'm just defining a strike price for a European option. The next line calculates in vectorized fashion. Again, no looping at all. In vectorized fashion I calculate for the 10 million end-of-period values. I calculated 10 million end-of-period payoffs of my European option. Single line of code, no looping. And last but not least, here the last analytical line is to get the mean of my payoff and to discount it back to today and done with my option price. So, rather sophisticated quant finance if you like and a total of maybe eight, nine lines, 10 lines of code to simulate 10 million end-of-period values for my stock price and to calculate the estimator, the Monte Carlo estimator for my European option that I have defined here. Financial data science. First thing we want to do and this relates to what I was saying with regard to open data is we need data to work with. And here I'm choosing data for the S&P 500 index. And again, it's more or less a single line of code. I'm using here a library which was previously in pandas which is now separated into pandas underscore data reader. And I'm retrieving data from Yahoo, open data and it's a single line of code and I have my data. So, this is kind of the story that I meant before. I had to go to my professor back then. I needed to pay 20 Deutsche Mark. I got back data that he downloaded from a source via a modem, put it on disk and I went back to my computer and started analyzing it with a spreadsheet. So, you have a single line of code, no 20 Deutsche Marks, no walking around, no disks and you have your data available. And this not only for this one thing, for kind of a million different things you can think of and I still find this fantastic. Doing the visualization, here you see it's another single line of code with pandas, just use plot. I'm providing here only the size of the figure as a parameter and you have the visualization for that. When we now read another source in for the big volatility index, kind of more sophisticated advanced index if you like, it is based on options traded on the S&P 500 and the implied volatilities are taken and some interpolation taking place and this gives kind of a forward measure for the implied volatility for 30 days. Same thing, works in the same fashion, data reader providing the symbol, the data source just picking out one column and I have my second data set. You can also visualize that. So, nothing new about this and you see here this is kind of a different time series. This is a mean reverting time series. Volatility is kind of nothing that goes to infinity in general, also not to zero. Only to zero if you have a stable and non-changing time series. So, two different time series but managed and worked with in the same way. What you then do, financial data science is you combine two data sets into one. This is easily done here with pandas. There are multiple ways and when I'm doing my pandas trainings, I'm presenting, I don't know, I think six, seven, eight ways of combining data sets, maybe nine, 10. But here I'm choosing the dictionary approach where I just say, well, the first column in my data frame should be called SPX and providing the data of the first thing that I retrieved, the second column VIX and so forth and together this should then form my data frame and again, this is a single line of code and I've combined my two data sets into one. This is, maybe in Excel, what you would do is kind of you copy something or there's one thing, you put it then in there and so forth. So, I'm not saying here that all this what I'm presenting is not possible with other technologies, hopefully it is but I'm emphasizing the point that it's quite convenient and easy to achieve the result. Now we have the data combined in a single data set and again, we can plot this now in a single plot. What I've chosen here is because we have a huge different scaling of the two time series. I'm just saying that the second time series for the VIX chart, please be put on the second axis and again, you see here in this part, there's always a single line of code and we get the result, there's not kind of big programming and doing crafting and even when you think of kind of Excel and moving around and getting this scaling right and here and opening the options that you need to do, I mean, of course, you have to learn that. You need to know that there is kind of a parameter which is called secondary underscore Y and so forth but if you're using this on a regular basis, that is kind of easy to master from my point of view and again, all what I'm doing here is in principle nothing to do with kind of proper programming. So you don't need to know, even to know about functions, know about classes and O or P or whatsoever, this is simply kind of from my point of view a meta language which makes it really easy to do not only financial data science but other data analysis applications as well. Now, two more combining algorithmic things with the financial data science wanna calculate the lock returns. This task is indeed accomplished by a highly vectorized operation. So think of calculating the lock returns for the two-time series. You have two-time series what you typically would do in other languages is kind of you would iterate over every single step. You would take today's prize divided by yesterday's prize and then take the logarithm. Then you would get tomorrow divided by today's prize, get the logarithm and so forth. Here, I'm having a matrix, a table-like thing with this data frame and what I'm doing is just take this whole thing, divide this whole thing by itself, shift it by one day in our case and take the logarithm of the whole thing. So it's kind of in multiple dimensions over multiple layers. I'm doing here vectorized operation and I'm ending up with the results object which is a data frame again and I'm having here obviously in the first line I have nothing because there was no precursor value but from the second day here, the 5th of January, 2010 I have my lock returns. And this again with a single line of code, no looping, nothing, I don't have to write code nor to do other sophisticated things. I can then, given my lock returns, for example, doing a little bit more analysis, statistical analysis here calculate by the call of the correlation method, the correlation between the two time series. Single call, again, I don't even need to know exactly the definition and so forth. Just call the method and get back here the correlation matrix. Rather simple because you only have two time series so we have two times two values with the ones on the diagonal, obviously. Last but not least is my last example. Again, a single line of code. This is now from the Seaborn library where I call joint plot on my data set and provide the parameter regression and one other parameter for the size. And you see what we get is indeed the regression through my lock returns where the lock returns are plotted with SPX on the X axis and Wix on the Y axis. We get the frequency distribution, the fitted distribution, we get the Bayesian aspects for the regression. So with a single line of code, we get done lots of statistics and plotting in the same step. And this is kind of, there is the quote from Tesflanders, 1911, use a picture as words worth a thousand words. There's kind of a, because I've worked intensively on this topic, kind of a study by ATAC University about this very fact that there is a negative correlation between the Wix and the, it's actually not about the Wix and the SPX is about the V stocks and the Euro stocks 50, it's the same thing. Similar index, similar volatility index and the empirics are the same. And here as a student or interested observer knowing Python, you can replicate the major results of this 50 or 60 pages study with maybe, I don't know, 10 lines of Python coding. Maybe with a much nicer plotting than you find in the study. Not bashing the study, but just emphasizing the point that it's so easy to replicate these things these days with open data or open source technologies and all the plotting libraries and stuff around our ecosystem that is available. Yeah, but it's nice. You might say, well, we see the benefits and this is now a little bit influenced by my experience being owner of a company or we're working with big financial institutions, smaller ones, medium sized ones and where we need to convince management and those who are responsible that there's kind of a good way to do it because if you think now again of the corporate world and the traditional IT model that we're coming from where the big IT companies provide their professional solutions based on a licensing model and the support model and a hotline thing and so forth, this is what they are used to. There is somebody coming in, the sales manager, they sell, they discuss the agreements for the licensing terms, they discuss the support contract, they give them the documentation and if something is out of order or there's a problem, they call the sales manager, they call the hotline, they speak to the technicians and so forth but how do you do all this with open source? More often than not, there is no company, there is no sales manager, there is no contract, there is no hotline, there is no nothing and we still, I mean, this mindset of open source is more and more getting to the corporate world but still we have many environments especially in the financial industry with the compliance issues, the legal departments and so forth, they said well, this is not really what we can live with so rather stick to the old model and buy something where we have people we can talk to and that we can sue in the end. If something goes wrong, then running the risk of open source but I think there are good ways of learning it and of doing it. The one is, of course, if you wanna learn something, you need resources and again, 10, 15 years ago, I think the first Python Finance book by Wiley was Financial Modeling with Python. This was kind of really the first book and it didn't sell that well because maybe it was too early, I like the book actually due to the fact that it addresses things that might not be the sexiest in the world but that are important in the finance world and it took then years and years after the second finance-related Python book came out but now we have, thankfully, we have a couple of them but usually what you do when you start learning a program is not that you start with a specific book, you rather try to master the language and with regard to standard Python books, there is not a wealth of books available. So I don't know how many Python books there are from O'Reilly, from all the publishers that you might know. I've picked out one because it's so close to what we do, to what I do, it's the book by West McKinney who has written a pandas and it's best known for it but it's also well known for his Python for Data Analysis book and I just read on the slides of one of his recent talks that he is in the course of writing a second edition I'm really happy about that. I think when this book came out, I think it was 2012, 2012 to today, four years later, so many things have changed in pandas, not that many things in the number world but so many things have come on top of what has been available in the pandas world that, fortunately, he is now updating the book. Then there's my book, so if you have mastered the basics of the language and know a little bit about finance, you might go as a next step to my book. So this is also selling quite well, we said all over the world, I think we might be close to maybe 10,000 copies of this book now, so I'm really happy about that. Then you get more specialized. Python for Finance is rather general, more or less having this hybrid focus on learning a little bit more Python, maybe against the background of finance. This is a real, I would say, a proper finance book illustrating everything in Python, so much more specialized, targeted towards a finance, a quant finance crowd indeed. And then with this book, which comes out in November, it's getting more specialized. So within this derivatives area and the focusing on listed volatility and variance, involved things and complicated, explaining how, for example, the v-stocks or the vixas that we've seen are calculated and so forth. So we see this specialization over time. 10 years ago, notebook, then the first came out and the first ones are usually are more general and now we are getting more and more specialized, which is, from my point of view, an indicator for the maturity of our industry. And just to mention one thing, which is even more specific, a book that I have on my bookshelf as well, that we have used to implement this model is about the Sabre-Saber-Liber market models and practice and you see it just in the sub-line with examples implemented in Python. Also really, really heavy on the math, on the model side, but using Python to illustrate everything that is done there. So from a financial education and practitioner perspective, we now have sources. And these are only five out of, I don't know how many Python for finance books in general are relevant. Maybe the number is already at 30, 40 or even 50. I didn't do a survey in this regard, but there are a couple of them available right now. Then of course, you need training. So the traditional business model, IT model has been, this is our technology. You sign the licensing contract and our experts come and train you. But for open source technology, there's obviously also a need and this is also today and I always like it like here when there's kind of a tutorial day or tutorial tracks at least, that you not only go to the conferences, drink coffee and eat things with lots of sugar in it, but that you can also learn, that you can sit down and do. And for my tutorial tomorrow, I really want you to write things, not to listen, but to sit down, take your notebook and write stuff. I will provide in the end all the code. But I think doing it is kind of, you can listen to music or you can play the piano. There's a huge difference. And if you wanna play the piano, you must play the piano. I recently read that, I don't know exactly how it was, is that you cannot have somebody else doing your pushups. If you wanna exercise, you have to exercise. By watching soccer on TV, you won't get any more fit. So just do it. Then we as a professional company are obviously offering things like in cooperation with Fitch Learning, we're doing the For Python Quant series. And we start, this whole thing started out as a conference only one day in New York, 2014. Then we added two days of training, then people ask, yeah, for an introductory class, then we added the third day and this year for the first time, we've had five days. From Monday morning with a meetup on Monday afternoon until Friday night with the conference and four full days of training. And I would say in relation, I mean, obviously you have more people at the conference, but in relation, the bootcams and the workshops are more popular than the conference itself. And we try to do our best and just discuss, or currently discussing to do this, not only in London and New York, but also in the near future in Asia as well. Then what I started out, because so many people ask me and it's really going like crazy and really successful. I wouldn't have imagined that we just started online trainings. And we are now in the first cycle, so to say. We are providing trainings about financial data science, algorithmic trading, computational finance and special topics. So if you're doing all of this together, then you can earn a university certificate. And this is something I'm even more proud of actually, that I was able to get this accreditation of our local German university where I'm living. And I say, well, this is, of course, fulfills all the academic requirements that we have. And if you now do our online training classes, which roughly takes kind of 12 weeks, if you do all of them in a cycle, you can earn a university certificate. Maybe not as interesting here in Asia or in the US, but in Europe, even you are able to get five credit points for what we offer. So it's kind of, it's indeed an official accreditation in a sense that if somebody is doing in the UK a master's program, we have in Europe this European transfer credit system, ETCS as it's called, you can earn five points for your master's program, for example. But apart from that, if you're pursuing a career in finance, it might also be kind of interesting to have a certificate not only by our small company, but by a German university as well. I'm really proud of that because I think we are the first, even in the world, who was able to offer out of a corporate context such a university certificate. And what we are doing right now, and this is the company, it's the HEW, Southern University of Applied Science. We are currently planning a master of science and financial data science and computational finance decree which is centered around open source and quant finance and what I've been showing is kind of the first step, so to say, and we will design a couple of modules which you can as a professional, for example, do separately, but in the end, if you have done all the modules, you can earn a master decree itself. So I'm really happy about that because this is kind of something, yeah. I always liked teaching and researching and now I can combine our professional work with the academic side as well. But what we also need to get, or a group that we also need to get on board are the researchers actually. So the professors, the teachers, the researchers, not only that it must be nice to do it here and there, but those who are influencing, and here you see Jim Gatherill, those of you who have a little bit to do with finance might know him for his book, Volatility Surfers, which is a very well-known book. He's also a well-known professor. He is a presidential professor at the Baruch College and the president for the master financial engineering program in New York, one of the highest ranked MFE programs in the world. And he is known again for his volatility research and when he gave a talk in May at our for Python conference, the keynote, I was able to convince him to convert his R code to Python, so this is something I'm also working on and we volunteer then to translate it. And he originally handed in a Jupyter Notebook, again Jupyter Notebook, listen to it, with Python code, but we translated it and he for the first time at our conference presented his whole research paper and the whole presentation, so to say, with Python code. We have to do work here as well to convince people at the university, those who are responsible at Python is kind of a good choice to do. Open communities, having a look at the time, you see here PyData London and I elaborated on the open communities already and similar to my book story where I started with Wes's book with regard to data analysis, which is general out of our PyData ecosystem, I'm starting here with the PyData London group, which is I think the biggest, maybe even in Europe, I'm not really sure, but they have 3,000 members. Let me enlarge that a little bit. Yeah, oh, three and a half thousand, I want to see it here, three and a half thousand members. They have monthly meetups with almost 200 people, a really, really active and you see here, this is hosted I think for the whole year again by AAHL, a hedge fund, so they're kind of this hedge fund, this is the finance guys are interested in the PyData ecosystem, obviously, so it's kind of a direct connection and you have the availability, they have always high level speakers with very sophisticated topics presenting there, free entrance, free beers and I guess snacks as well and you can get the thing firsthand, so to say and obviously, again, many, many people out of the finance base there, but I would say in my meetup group, which is roughly half the size we have, this was from two days ago, we have roughly 1500 people and maybe only seven to eight meetups per year, with the big sort, but still, we are a niche and a niche, so we are niche squared or even cubed, if you like, and that we do Python for Quant Finance, not for retail affairs, Python and Quant Finance, but we also have talks about R and Julia, I see rather open, we even had a five talk, I guess for a five talk micro-conference about open source and Quant Finance, I also did an open source for Quant Finance and Frankfurt conference full day, so we tried to be as open as possible not only Python and Python only, and you see you have lots of opportunity, at least, and in London, I'm doing the same thing in New York as well, but I'm not as open in New York as I am in London to organize that and it's a very vibrant and an active community there. Open source libraries, just a quick run through if you're not aware of it, but interested in it. Pytholesians, one example for doing trading, backtesting, interacting with Bloomberg, written by a friend of mine, we probably will do two weeks from now a public open webinar where he will present via our platform the library, so very interesting. PyIgotrade, something we use in our Igo trading class to illustrate backtesting, really, really nice, very sophisticated, can do many things. Open source, you can have a look at the code, if you want to adjust it, feel free to do, or to contribute even better. Then we have SIP line from Quantopian, so if you like, this is, I don't know if they would see it the same way, but you could think of Quantopian being a backtesting platform and open sourcing the backtesting library underlying it, you could say, well, they're open sourcing their secret source more, that's because this is the model that they have, they have the platform, but they see it different, they see benefits out of open sourcing it, and obviously our community can only benefit by that, and here, one example, I didn't provide a screenshot of GitHub, you hear code, code posted on GitHub, and if you are interested in how they did it, just have a look at it, so you can learn it, you can adjust it, you can pick out parts of it, or better even, again, contribute to it, make it better, I think there will be more than happy. Our analytics library, same thing, I remember me sitting in New York on Long Island at the university when I wrote my PhD, not there, but there was visiting there, and I read for the first time a certain paper, and I was really amazed by the theory, risk, neutral, pricing, and general, for me, it solved the whole financial problem of pricing anything, so this was for me, the most general thing I ever encountered, and I thought, well, if this could be just implemented technologically, this would be the best thing in the world, and from my point of view, this implements the risk, neutral, pricing, or library, the risk, neutral, pricing theory of Harrison Krabs Pliska invented at the end of the 70s when nobody ever thought of anything like a cloud or whatsoever, so fantastic what you can build even with small teams and what you can reach using the technologies. On the other hand, I mentioned Quantropion, and they're not only outsourced, a sipline, but also they're recently implemented Pyfolio Performance Management Library, so really happy that they are continuing to work in this open source, outsourcing stuff, and people working there, like Thomas Wiecki, a friend of mine, is also a German guy, he contributes also to PyMC3, which is a Bayesian statistics library and so forth, so this whole universe that you see here is getting more and more open, and I'm happy to see these models and technologies getting more and more important. Our platform is more or less, if you like to say, just a collection of tools, so with many thousand people who use Jupyter Notebook, if you don't wanna install Jupyter Notebook, just go to our platform, pqp.io, you can register after maybe 30 seconds, you are able to use the examples on the platform to just try out Jupyter Notebook, but you can do many, many other things. We are mainly using it these days for trainings where we host the resources, where we put the videos, where you have kind of a training area and so forth, this is our main usage, but a couple of companies have also licensed the platform for internal purposes, but we see it more as a delivery mechanism with what we do, and then there is, I'm really thankful, the professionals, the targeting, the upper market, if you like, and when I spoke of Bank of America, Merrill Lynch with 40 million lines of code in production for the Quartz risk management system, the guy who was responsible for that over the years founded Washington Square Technologies with another guy who was responsible for Athena and JP Morgan. So the two heads responsible for probably two of the biggest financial applications built on Python in the world, have founded this company and are now providing for small and small and medium-sized companies the same approach, and they are also referring, it's hard to read, I guess, here, they're referring to this, you see they have been working at Goldman Sachs, JP Morgan and Bank of America and what they try to replicate. Again, it's completely Python, completely maybe not, but it's mainly Python-based, what they provide and it's used already by a couple of insurance companies and others for institutional risk management and pricing. Last but not least, and this is just a wrap up, this is my final thing when it gets to how to do or how open source and all these technologies influence financial education and financial practice and what we try to do and achieve with what we do and yeah, the example is about derivatives analytics with Python, you've seen the book already and it starts from my point of view with the book. And the book itself started with my university lecture about numerical methods for market-based option pricing, now published by Wiley and it's not only that there is a book which is kind of something we have since Gutenberg available quite a while. These days we have other means and mechanisms that we can use to spread the word and to make use of the ideas behind it. Obviously I have, there was one, I forgot to do a break here, so I have to show it that way. I've hosted all the codes, 5,500 lines plus Jupyter notebooks on a GitHub repository. This is the first thing. You can go, you clone the GitHub, you have everything. This was back in the day, we had books, then they provided books with CD-ROMs, some of them at least. Now obviously no CD-ROMs needed anymore. You put it on GitHub and everybody was a little bit capable of doing these things, can clone and use the stuff. Otherwise probably they wouldn't read the book. So this is kind of a nice and easy way and free, cost-free way of doing it. The next step, and this is how we use the platform as well, the one thing is provide the code, but how can I make sure being not a professional, pressure is not the right word, be not a commercial solution provider with a big support team, how can I make sure that people can use it easily? Yeah, we can host the code. We can provide the execution platform, not only the code itself, but the place where you can execute it. And this is what we do with the platform. You have a separate platform, you don't need to get to the GitHub repo, you get it on a platform, you lock in and you use it immediately. So we provide you even with the power, with the capacity to do your derivatives pricing things here with the library. And then you might say, well, nice, now I learned about it, but I wanna apply the techniques and the models and everything that I've learned here in practice. And this is where the library comes into play. Everything that is presented in the book is implemented in the library and much, much more. So all the derivatives and analytics related things that you find in the red book and then the blue book are implemented in a, I would say, yeah, production ready fashion in the library. You don't need to get from the functional approach of what built your own classes. This is done with DX analytics. And you see it here. I also mentioned in refereeing, so to say as a documentation for the library, the book. So you can say the book is documentation for the library and the library is an implementation of the things presented in the book. So this will be the next step if you are not an academic, but maybe someone working for a medium-sized hedge fund. And last but not least, this is what we started recently because I was asked for that for many, many sides is that we now offer this online training where I say, well, I understood this in here and there, but there's no lecture in my university or we don't know at our company. So we are providing around all these topics, covering the library, the math side, the Python side and everything in one of our online trainings where you then learn about NumPy, Panthers, option markets, risk-neutral evaluation, 48-base option pricing, and all the things that are presented in this book. So this is kind of a complete package that you can get all based on the technologies in our universe that we now have available. My final slide, Python for Finance, my summary. From my point of view, Python has become the English of programming languages, and parentheses for finance. Many other places as well. When I'm in London, many people, and I've never heard this in America, actually, in the US, but in London, many people have said it to me, well, it's really a shame that we speak English because we don't have any incentive to learn another language. All people who are visiting us, they speak English, where we visit other countries, most of the people speak English as well. And today, if you learn only one foreign language, most probably it is English. And from my point of view, and this is what I'm saying when people are asking me, if you have only time to learn one programming language for finance thoroughly, then learn Python. Of course, if you're a specialist or genius, you can learn others, but from my point of view, Python is the one that you should choose if you have time constraints and if you want to be as flexible as possible. And I'm pretty sure, and I'm getting asked this question over and over, Python is there to stay, not only for the next three or five years, from my point of view, at least for the next 10 years. And, yeah, Python will and has already heavily influenced how finance can be learned and can be taught and how finance is practiced these days. Thank you. Any questions? We can take one question. There's one. People have a good question. So, if you want to implement proper pricing and analytics and risk management in Python, eternity for the speed with files that actually constraints, because what do we do if you're in Python or if you're used to it from today, you're likely to see that first or see it or something else, what do we do? Here, this is also a regular question, actually. I think we are getting more and more flexible with regard to the options that we have. Of course, if you're doing pure Python, as an interpreted language, when we speak of C Python, the standard implementation, the typical deductive reasoning is Python is interpreted when the interpreted language loops are slow. If we implement financial algorithms, they are loop heavy, so the implementation might be slow. Therefore, Python is not suited to do finance, but I have covered already one solution in the sense of that when you use NumPy or Pandas, you can easily vectorize your operations and then you get the benefit of using the performing implementation under the hood of C or in some places, even still for tribe. In addition to that, you have many performance libraries that you can use. Just to mention one number, number is a dynamic compile library, which makes it possible especially to compile loop structures, nested loop structures written on NumPy. So what you do typically is you take the NumPy infrastructure, the NDAray object, and you write your Python loops on top of that, just to provide the compiler with lots of information. And there you can see speedups sometimes of 50 or 100 times. If you then go and wanna make it even more flexible and more general, you can use Thyson. Thyson can, as a static compiler, can understand the interpret compile in that case, pure Python. But what you can add is all the nice things from the C world that you can do static type declaration. You can use C structs, you can use C libraries, for example, for the number generation or whatsoever. And you have kind of a flexible spectrum, like kind of a fader that you have. 100% pythons, not enough for me, I go to maybe 20% C, 30% C. Even in the end, you can completely, as they say, thysonize your code and you end up with something statically compiled, which might be exactly as fast as possible. And when you then think of that, typically the bottlenecks, if you have 1000 lines of code, you have bottlenecks, which might be 10, 20, maybe 50 lines, maybe 100 lines of code, but usually not more than 10%. You pick these out, you thysonize them, you dynamically compile them, and then you get the speed of other technologies as well. And you have the convenience still left of doing all the logistical things in Python. IO, for example, is always high, not always, but most of the times with non-py panes and so forth, high performing with Python. There is no overhead or whatsoever involved. So I don't see, speed is not an issue anymore. And still, not speaking of being able to parallelize things and being able to resort to C++ and interfacing with C++ as well, or to this end with any other technology. You have all the options, but starting with Python, you still have the efficiency and productivity that all these parts of the ecosystem bring with them along.