 As I was preparing my slides for today's presentation, I was reminded of this meeting that I had almost exactly five years ago on this Stanford campus that changed my life. At that time, I was about 15 years into a career which was marching down the semiconductor process node starting at 0.65 microns all the way to 28 nanometers. And it was getting a little bit repetitive and boring. So I sought out the advice of one of my longtime mentors and advisors, Professor Abbas Elgamel in the electrical engineering department to see what I could do next to get some more fun and excitement back into my working life. And he told me, oh, go take a look at Smart Grid. It's really interesting. And so I'll be honest, at that time, apart from taking a couple of undergraduate courses in India on power systems and cursing the system while spending some time with my kids and wife in India in 140 degree heat in June without power to power the fans, I had never really thought about this industry and the fact that I might actually do something in this industry. But then when I started doing some research, I realized that this amazing infrastructure which is around us was undergoing a transformation of its lifetime. There was just incredible amount of innovation happening on all fronts, whether it was on the technology side in terms of solar and wind and electric cars and storage, or on the pricing side trying to change the behavior of consumers, or on the competitive market side where both wholesale and retail markets were getting deregulated. There was just a huge amount of new stuff happening. And it sort of became obvious to me at that time that all of this innovation will not happen unless every player who is trying to innovate gets a very deep comprehensive understanding of how all these pieces are going to fit together and how this whole system is going to continue to work. And furthermore, if we can actually build that understanding that will give us the opportunity to optimize the system and eliminate some of this inefficiency that fundamentally stems from a lack of understanding of the system. And so as I was thinking, I also realized that there are some incredibly complex systems that we understand really well, and we have optimized every last drop out of them in terms of how much power they consume or how much time a signal takes to travel across these. And these were the semiconductor systems which I happened to know reasonably well by then. And so I thought maybe I can use some of the ideas that have been developed over the years in this domain and apply it to the power systems. And as it turned out, fortunately for me, the laws of physics remain constant. And so even though these chips were tiny, the signals were traveling at very high frequencies and we had understood how to model these systems that can have tens of millions of transmission lines and optimize these. And so I thought, okay, well, maybe we can apply this to the power systems. And then I had my second good fortune which is I got introduced to Steve and he used his magic and created conditions which allowed me to spend some time on the best place that exists on the planet Earth which is the Sanford University. So I got some time to spend here, work with some incredibly brilliant people, Balaji, Ram, a lot of you in the audience here to just refine some ideas. Bob here, I can't name everybody, but you know who you are. And so that's how I started on this journey. And today at AutoGrid we are about 35 people, still very early stage in our life cycle. We raised about $20 million in venture capital. We have about 15 utilities that are using our systems across the world, mostly in North America, but we have a few customers in Europe as well as in Asia today. So my advice is that if you are thinking about getting fun and excitement back, to spend as much time as possible here and get to know Steve. So that, let me get to my talk. What we are doing at AutoGrid is building on top of what has already been happening over the last several years. So over the years, I think there is more than 100 billion that has already been invested in upgrading this entire electricity supply chain all the way from generators to the rooftop units that might be running the AC system in this building. And there are, okay, perfect. And there are a number of very, very successful companies, companies that have built sensors ranging from smart meters and data systems as well as data management companies that have built systems on top of these sensors to collect and manage this layer. But as a consequence, what has happened in the industry by and large is that there is a fragmentation in terms of how many systems are out there and where the data is sitting. And there are these hundreds of different data silos that have been created within a typical utility or electricity supplier, which contains pieces of this data. And to be able to get a really comprehensive understanding of the overall system, one needs to bring all of these pieces of data together in one place. And so that's what we ended up building at AutoGrid. We call it the energy data platform. We take data from many different sources, any source that's available within a utility that can include smart meters, but also include sensor data. And we use that to build applications that utilities or large customers of these utilities can use to optimize their energy usage and improve their overall energy productivity. And there are a number of things that are really common across these applications. One is the whole notion of big data, the volume, the variety, the velocity of this data, the fact that you have to bring this data in real time and process it. But the second thing which actually Ram alluded to is also a deep understanding of the underlying physics, how the electrons flow on this system. And if we can understand both the data aspect of it and tie it with the physical understanding of the system, we can design algorithms that can predict, that can forecast what's going to happen on this system and take control actions. And so we have built this engine, which we call a predictive control engine, which is able to bring these two domains, which so far have been very different in the industry. The big data folks typically don't do high performance numerical computation in the same system. And by doing that, we can actually now do closed loop optimization. And once we have this, it can be reused across many different applications. We can do demand response. I'll talk a little bit about that. We can use how users can reduce their electricity consumption or electricity cost. We can see how we can do better voltage optimization on the grid as well. So I'll just use two examples in this talk. These are real life examples, live deployments that we have out there with customers and they're like today, actually yesterday. I'm not using that example, but we did our first event for the summer at City of Palo Alto utilities for this summer, which has been our first customer, which was also courtesy Stanford and Steve that we got that introduction. And so the first application is called demand response, optimization and management system. This is something that we introduced in the market in 2012. The basic problem here is something that I'm already talked about. About 20% of the electricity is consumed for less than 2% of the hours. But because there is very minimal amount of storage on the grid, we have to still build the infrastructure to keep on standby so that we can serve these 2% of the hours. Which if you aggregate over the entire US is about 400 billion in cost. And so if we can eliminate that, that would lead to 400 billion in savings. This is not something which is conceptual or researchy. In US alone, in 2012, 22 gigawatts of demand response was actually called to shift the peak or used as a generation. And just to give you a context, 22 gigawatts is about 25 San Onofre nuclear plants which were offline just because of these type of programs. So this is working, it's real, it's here, it's working at scale. There are markets which allow customers to go in and bid this type of demand response. But also, if you look at the total potential, this is a number from FERC. We are still sort of looking at only the tip of the iceberg. There is a huge amount of potential on how this can have much deeper penetration and we can get to about 20% of overall demand served by these type of demand response programs. But what's really preventing this from happening? I think the key issue has always been this disconnect between what the customers want, the end customers, the users of electricity, and what the suppliers, the utilities want. At the end customer level, the customers are interested in maintaining flexibility and control. Only then would they go and enroll in these programs. And so they want to be able to participate when it's convenient for them, when nobody's in their house for a party or a visit and opt out whenever they want to opt out only if the prices are at the right level. But then the utility wants to use this resource as a generation. They want it to be dispatchable and they want to be able to press a button and get a response that they can count on. And fundamentally these two things are really orthogonal to each other. But that's where big data comes in. That's where the power of analytics comes in. Using analytics and using some of the ideas that Ram actually talked about in his earlier talk, we can get the best of both worlds. We can actually create these programs which allow our customers to behave as they normally would. They can opt out whenever they want to. They can have random behavior. They can live in Brazil or in Germany. But, and so they don't have to move. But they, but at the aggregate portfolio level, we can still get the reliability that is at par or better than a generator. And that's really the key sort of insight that we get from big data because we can analyze the past behavior of every single user and we can build these models that can predict what these users are going to do in the future at an aggregate level with pretty high degree of certainty. And furthermore, we can also predict what type of actions they are going to take when you provide them different types of incentives. And that's really the heart of this issue I think. So this is a program that has been running at Oklahoma Gas and Electricity. We partner with one of the largest AMI vendors in the world, namely Silver Spring Networks for this program. Oklahoma Gas and Electricity about four years ago decided that instead of building a plant which would cost them about $350 million and will be used for less than 50 hours during the year, would implement a program where they would give an incentive to their customers to change their usage. And a customer who would participate in this program would normally get about 40% discount on their electricity bills instead of paying 10 cents, they will pay a six-cent price on most of the hours in return for them agreeing to receive a price signal which could be higher during a few hours in the year when the grid is stressed. And so this program has been running. The net net of the story is that they have deferred the creation of two-peaker plants. There are about 80,000 customers who are enrolled in this program who get a daily price signal. This price signal can be one of five levels on every day and it could also be another critical price that could occasionally happen for them. On an average, a customer is saving about $190. They have won almost every award that is out there in the grid area for this program at this point. They have been running ads across, if you go to Oklahoma City, you will see radio ads and billboards and after thunder, probably this is the most interesting thing going on there. They, there is in one of the ads, there is a grandfather who is talking about the value of this program and he says that, well, this means that I can see my grandkids one more time during the year and so it's very appealing and the satisfaction rates from the customers are well over 98% in this and yet they can get the reliable resource that they need from this program. So this is the loading order. Last year they achieved about 70 megawatts of load. They have a number of large turbines and their smart hours program actually shows up as one of the generators in their dispatch order here. We are able to take data from their system every five minutes in real time and we can tell them what's happening on their network from all of these customers and we can forecast every 15 minutes what's going to happen on the grid for a given price signal and we can update this as the weather changes but as the data comes in and I have some plots and given that they have many different price signals, you can see a very clear price elasticity here which was nice to see when we actually analyzed the data. So as the price goes up, you see the reduction actually becomes better. If you look at the value of this program, there is clearly an economic value which is the capacity value that I talked about by differing the peaker plants but there is also an energy value which is the fact that you don't have to run a generator and then there is of course some value in terms of providing some balancing and so on in the grid. And so we did some calculations and in addition to the capital expense by actually bidding this energy into the market, OG&E can save an additional million dollars a year for this program. And key to this is actually a forecasting engine which allows us to really model every single user and tell what's going to be available because unless you know what's going to be available, you really can't bid it into the market or try to monetize it. This is another chart which shows why it is important. So here I'm just doing different days and looking at the range of what's the variation between the load shed on different days. And so on an average, the load shed is around 45 megawatts but on a particular day on the same program for the same price level, it can be plus minus 20%. So if you are not doing a very precise analysis of how much load shed you will get as a utility or as a service provider, you would not really know where to, how much to bid. And if you want to be conservative, you will end up bidding potentially here and leave some of this on the table. If you are aggressive, you'll bid here, but then you will have to pay penalties at this point. And so by doing this type of analysis, we can shave some of those margins and make more money for our customers. The second example is actually on the other side of this equation. So what happens to customers? So if you look at a large customer, a typical commercial building like this, there are two components to the electricity cost. One is the normal per unit cost that we pay as residential customers. But what a lot of people don't realize is that there is a huge piece of this cost which is based on what is the peak usage in that building during a given period of time, typically a month and sometimes it could be during the year. What that means is that even if I'm using very, on an average, I'm using very little electricity in a building, but for 15 minutes in a month, I peak, then I'll have to pay a very, very large premium for that. In Palo Alto, the premium is about $20 per unit, while the average price of electricity is less than 10 cents or around 10 cents per unit or $20 per kilowatt. And so if you look at an average, overall, within California, about 40% of the total electricity cost happens for usage in less than about 10 hours during the month. And so this is another curve. It's a low duration curve where you can see, this is a real building here in Palo Alto, actually, where the peak usage, if they're able to reduce their peak usage by about 5% or 10%, they can save roughly 75 to 100K during the year. And to be able to do that, they have to change their behavior for less than three to four days during the month. This is a building complex in Redwood Shores where our office is. That's a set of eight buildings. They're paying 1.2 million per year in demand charges. And again, if they want to curtail 10%, they have to do something only two and a half times per month for a building. And so a very simple incentive or program that we have done is we can forecast when these peaks are going to happen because we know what their usage in the past has been. And we send them alerts. And these alerts could be more or less sophisticated. On the simplest level, they can be just emails or text messages. And then if the building has some sort of automation, they can receive automated demand response signals using protocol such as OpenADR and take action without a human in the loop. And so we can send these type of signals where they don't have to wait for their utility to tell when it is convenient for the utility to run a demand response program. It's almost their own personal DR program where they are reducing their own cost. And the payback is immediate because they don't need to do anything new to take advantage of this. So in our building, they usually turn off their water fountains which are in the common areas or they turn off the external water fountains lighting in common areas. They may pre-cool a building so that it can run in a more relaxed manner during the peak hours. And these buildings can save substantial money without additional capital expense. And so that's just the same concept. I mean, they get the system. We do all kinds of analytics on which messages are effective. Some people like to save money. Some people like to be environmentally conscious. We can do A-B testing around it. We can see what type of frequency do you have to send reminders on? Are the day ahead notifications good? Do you have to tell them during the event that they should do something? One of our customers is turning off their EV chargers to a lower level of charging when the demand chargers are out there because that's directly adding to their overall cost. So these are just two examples of how big data can be used. I can obviously talk more about some of the other examples during the panel but clearly today in the grid, these programs are real. They are working customers like them contrary to some reports out there and they are saving money both for the end customer as well as for the utilities and making the overall system a lot more efficient or a lot more productive. Thank you.