 So every day we produce massive amounts of digital data through our browsing history, transaction records, GPS location, cell phones, sensors and all this big data analytics has disrupted large sectors of the economy and it's making the fortune of Silicon Valley companies. And here in San Francisco we're the heart of this big data revolution. We have the greatest tech talent on earth working to make our lives a little better. And they're helping us make new professional connections, figure out what books we might like to read and what's the most efficient way to avoid Friday traffic on the Bay Bridge. And this is all great and it's making our lives slightly more comfortable. But is this really all we can do with this great technology? Is there a way we could use data science to actually create social impact? So first of all, what is data science really? It's this new field at the intersection of math and statistics, computer science and domain expertise or operational research. And it's really about three things. Knowing how to ask the right questions to the data through these new amounts and types of data we're generating, using statistical methods to answer those questions and then software engineering to implement a solution and build data products that are actually going to be usable by others. So in the private sector, businesses have invested millions of dollars in these techniques to figure out, to know their customers better and predict their tastes and preferences and design solutions that match those needs. And the question is really, could we use these same techniques but instead of predicting our likelihood to buy stuff, could we use them to identify people who might have a high risk of health complication? Could we use the algorithms that Uber has to optimize ambulance dispatching and reduce wait time? Could we do what banks and entrances are doing to build your risk score? But instead of that, predict who might be, what are the people who might be at risk of being stuck in poverty traps? And the thing is, if we invested as much in knowing the beneficiaries of social programs as marketing departments do in knowing their customers, we could know how individual might respond to specific social interventions and we could design, we could match them to the most effective social programs and we could tailor social programs so that they're overall much more performant and we could increase performance dramatically and reduce operational costs. And so we know that all of this is possible with data science. So the question is, what does it really take to do data science for social impact? So there's some of the tech enthusiasts or utopians out there who think all it really takes is just a bunch of smart data scientists who can hack the social sector. And famous investor Mark Andresen has this quote about how software is eating the world. And so just like it's overturned large traps of the economy, it might just take over the social sector to rationalize and optimize it. And then you have the skeptics or realists, depending on where you stand, that think that, well, eradicating world poverty and hunger might just be a tiny bit more complex than this. And mission-driven organizations exist in the first place to correct some of these systemic inequalities and correct market flows. And so there's no reason to think that just releasing the technology out there will solve any of this. And I have to say, I was a bit of a skeptic myself. I worked for several years at the Frontlines of Development and working with poor farmers in Colombia who had been displaced by the war, I saw firsthand how the state was unable to deliver on its promise to return the land to these people. Not because there was a lack of political will, because they had no idea who owned what. And collecting the data and processing those claims just took forever and ultimately failed. And if there's something that code is really powerful at, is automating things on a large scale. And that's when it clicked for me that sometimes when the conditions are aligned, technology can really be this instrument that carries policy and research and good intentions towards creating tangible social good. And that's what I came to look for in Silicon Valley and I now have an office with pickpock tables and I use the word disrupt in my presentations. So on one hand you have those guys, the technologists, who can really, really powerful stuff really quickly but they don't necessarily know what for. And then the development and the policy people who know a lot about the problems we're facing but they have no idea what we can do with data science. And so bridging the two takes time and effort and that's why we created Bayes Impact. We're a nonprofit that builds data science solutions for the social sector and we invest heavily in finding the most pressing social issues that also have the highest opportunity for technological leverage. And we build open source, data driven software that help governments and nonprofits optimize what they do in the long run. Essentially we're working on bringing the latest data science technology from Silicon Valley to the places it hasn't gotten to. One area that we think where we think data science can be really impactful is agriculture. GDP growth from agriculture is twice as likely to reduce poverty as any other sector in the economy and agricultural productivity is important because the global demand for food is expected to double by 2050. So essentially improving agricultural efficiency is a question of poverty reduction but also food security and building a sustainable food system. So this is Don Andres. He's a rice grower from the Tolima region in Colombia. And over the past few years, his average crop yield has declined from 6 tons per hectare to 5 tons per hectare due to high climate variability. Agronomic processes are really complex. Crop development is influenced by countless interrelated variables and increasing climate variability is just making this even more complicated. So for Don Andres to adapt to these conditions sometimes feel like a bit of a stab in the dark. But imagine if we could use machine learning techniques that identify patterns in the data and learn from historical cases to help Andres know what are the optimal steps that he could take to optimize his crop cycle. What if he could know what's the seed variety most suitable for his field and what's the planting date with the highest expected return this season? What if he could know how to prevent the spread of a disease before it even reaches his field? All of this we can do with data science and that's why we're building a decision support tool that collects data in real time and uses predictive analytics on soil data and climate data and harvest data to give Andres and smallholder farmers personalized recommendations on how they can optimize the entire agricultural cycle. And it looks something like this. They can receive real-time information about top crop and seed variety for their specific field and agro-climatic conditions information about when they should start irrigating how much fertilizer they should use and the pilot tests that we're running in Colombia show that we can get up to a 20% increase in productivity. So of course, all the big corporations like Monsanto and John Deere are racing into this analytics business and they have huge conflicts of interest. I mean, they're the same companies selling the seeds and the equipment and farmers are concerned about the fact that having so much data might enable them to tweak the market in a particular way. So that's why we're building an independent platform accessible and affordable to smallholder farmers in the developing world. And this is something that could be super powerful to reduce the yield gap and equalize the effects of technology. So there's a few aspects why I think this could be really powerful. One is that it shows how data can be tied to decision-making in a very direct way. So we're not just collecting data to write a report that's going to sit on someone's desk here. We're putting it to use in a very tangible way. And that's why at Bayes we focus on building decision-support tools that can bridge the gap between data collection and action. And with the rise of inexpensive smartphones out there we can now get a phone for $40 and this next billion people that's going to come online in the next five years there's huge opportunities to drive behaviors not just at the case worker level but all the way to the end user. Another reason why this is powerful is that it automates the data collection process. So instead of having to send surveyors to the ground we're now generating data directly by the users using the product. And so it's a much cheaper and more reliable way to generate data and this data we can now hand over to the state so ministries of agriculture can have an idea in near real time of what's going on in the agricultural sector and they can use this to refine their policy process and calibrate their agricultural subsidies for example. And finally it's open source. What does it mean to be open source? Well first of all it means that it's open. There's no gatekeeper, there's no one channel for progress so any community can use and modify the source code and adapt it to their particular needs. It also means that not having proprietary software avoids being locked in with just one vendor so you can change solutions as you wish. And it also means that it opens the door to having a non-profit organization or a public entity managing or owning this data and setting all the right regulations so that the data is not excessively privatized and we can make sure agricultural data is being put to good use. So I want to end taking a look at this picture at this cover on big data from the economists and the reason I like it is because it's pretty clear how challenging it can be for the human mind to make sense of the multitude of signal and the data that we're generating. But it also shows that ultimately there's a human being here holding the tool that turns data into insights and he's holding it and transforming it into a fertilizer to grow this really nice plant and not feed a dragon, for example. So instead of letting software and technology eat the world, I think we can shape it to help us feed the world for the better. Thank you.