Energy for Growth Hub
Multimedia Mar 23, 2022

Episode #11 Jay Taneja: Using Machine Learning to Inform Energy Access Solutions

Taneja, Professor of Electrical and Computer Engineering and Computer Science at the University of Massachusetts, Amherst, discusses his…

Better Metrics and Data

Taneja, Professor of Electrical and Computer Engineering and Computer Science at the University of Massachusetts, Amherst, discusses his work on creatively accessing data in data-poor environments, the challenges in electrification via grid connections, and his work as an engineer and professor in the development sector.

Jay Taneja is a Hub Fellow and an Assistant Professor of Electrical and Computer Engineering and Computer Science at the University of Massachusetts, Amherst, where he directs the Systems Towards Infrastructure Measurement and Analytics (STIMA) Lab. Jay previously worked at the IBM Research – Africa Lab in Nairobi, where he founded and led a team of researchers focusing on using analytics and mobile systems to improve electricity reliability and access. He holds a PhD in Computer Science from the University of California, Berkeley.

Show Notes

Amped Up:

Jay Taneja Interview:







KATIE:  I’m Katie Auth in Washington.


ROSE:  And I’m Rose Mutiso in London, and this is High Energy Planet, the podcast from the Energy for Growth Hub about new ideas to solve global energy poverty.


KATIE:  On today’s show what do nighttime lights photographed by satellite tell us about electricity reliability in Africa?  We talk with Jay Taneja, assistant professor of electrical and computer engineering and science at the University of Mass Amherst about how he’s found novel ways to research energy access in places where conventional data is hard to get.


ROSE:  We’ll also ask Jay how an engineer like himself came to address human well-being questions that are usually posed by social scientists, why he’s recruited and mentored dozens of African students at UMass, and the big unanswered questions he’d like to tackle next to push quality of life forward in Africa.


KATIE:  But first Rose and I get amped up about critical minerals in the clean energy space.


ROSE:  All that coming up on this episode of High Energy Planet.




ROSE:  It’s time for Amped Up, when we talk about what we can’t stop thinking about right now in the energy and development space.  And this week Katie and I are obsessing about critical minerals in the clean tech industry and what that might mean for Africa.


KATIE:  So when I was preparing for this segment I will say I was probably naively shocked at just the sheer scale of what we need.  So if you think about copper, for example, in just one electric vehicle there’s 180 tons – sorry – 180 pounds of copper, and 4.7 tons of copper go into any typical wind turbine.  So we need a ton of this stuff.


ROSE:  No pun intended there, lots of tons and pounds.  Yeah, I mean Katie, the numbers are mindboggling.  And the (IEA) is now projecting that demand for these minerals for clean tech could increase by a factor of six over the next couple of decades as part of this kind of global transition to net-zero.


KATIE:  Yeah.  So let’s think about the countries in Africa that already playing a major role in providing these minerals.  The DRC is a great example often cited.  It produces about 70% of the world’s cobalt supply, which is critical for battery technology.


ROSE:  Totally right, Katie.  And I think Africa is finding itself in the middle of this kind of geopolitical rivalry given the mineral wealth and its strategic importance for these materials.  So, you know, China, which already controls major chunks of the critical minerals supply chain has been eying African minerals to extent its dominance.  On the other hand, the U.S. is justifiably getting anxious about Chinese dominance in the space and is looking to Africa to help diversify away from this over-reliance in China.  So a lot of fun in games and Africa in the middle.


KATIE:  Right.  And at the same time there are African countries that are trying to exploit this on their own behalf.  DRC and Tanzania are both actively working to scale up their mining industries and kind of capitalize on all of this interest.  And there’s a lot of really optimistic takes out there about how developing these resources could really jumpstart development in some of the poorest countries in the world.  But I’m curious whether you think there’s a risk that that kind of development would just further solidify African countries’ reliance on raw material exports, and how, if we can, can we avoid that?


ROSE:  I don’t know.  You know, that’s the really tough one, Katie.  I mean, mining for critical minerals in Africa is already rife with all sorts of serious problems.  I mean, just look at DRC cobalt and the associated buyer lens, human rights issues.  (Inaudible) you see pictures of, like, little children in the mines.  It’s pretty bad.  So, you know, looking forward there are some transparency in (inaudible) governance initiatives that are already in place that could, you know, be strengthened and deployed in this next phase of critical minerals growth.  You know, it would also be great to see more African countries play a bigger role in mid- and downstream parts of the value chain.  So not just, you know, pulling out of the ground and sending (inaudible) abroad, but, you know, building more refining and processing capacity.  But, you know, I’m not going to lie, there’s a lot to be done if we’re to avoid falling back into old patterns.


KATIE:  And not just the supply chain for the minerals themselves, but it would also be great to see more manufacturing of clean tech.  And that’s something that development finance I think needs to adjust to.  We’ll continue watching this issue.  If you have an energy or development obsession, good or bad, tweet it to us at @EnergyforGrowth and we’ll include it in an upcoming episode.  Coming up we talk with Jay Taneja about how new data might fix the old problem of energy access and reliability.




KATIE:  Jay, welcome to High Energy Planet.  It’s so great to have you here.


JAY:  Happy to be here, thanks for the invitation.


KATIE:  Of course.  So we wanted to start out by pointing out that you have a great reputation for creatively using data to look at energy and infrastructure questions in a part of the world, predominantly Africa, where data can be notoriously hard to get.  For instance, you’ve looked at nighttime lights as indicators of electric power instability.  You’ve measured road quality through deep learning and satellite imagery.  How do you make those types of connections, and what are you looking for in data sets that help you pull out those relatively unique insights?


JAY:  So thanks again for having me.  You know, our group focuses on research.  We’re trying to uncover new applications, and a particular part of research is looking to see where technology has changed.  So when it comes to new data sets, new satellites going up measuring either at higher resolution so we can see smaller and smaller objects on the ground, or different satellites measuring pollution, or measuring nighttime elimination, we have opportunities to see infrastructure, to see the world in ways that we haven’t seen it before.  And so we’re always trying to understand what more can we do, what new can we do as technology tools get better?  Whether it’s the better imager, or able to measure a new concept, or whether it’s our algorithms to process those data so that we can look at more places or over a longer period of time more easily.


ROSE:  Okay, so Jay, following up on that satellite study, yes, the data in these regions is limited, but we already know intuitively/anecdotally/you and I lived in Nairobi – so anyway, we know that reliability is poor.  So how could your new data help actually fix the problem?  What’s the next step?


JAY:  So this is an opportunity where, yes, we have a qualitative set.  We have a rough idea that reliability is poor.  Part of the old management maxim is if you don’t measure it you can’t manage it.  And with reliability we’ve always had this rough sense that it’s poor living in Nairobi, it was always, oh, the rains are coming, we’re going to have really poor – we’re going to have a lot of outages coming up.  But without knowing where those outages are happening, how long they’re happening, whether they’re happening in certain areas more than others, how do you make investments to improve that system?  And so the maxim for what our team really works is how can we improve measurement?  How can we better understand not just where infrastructure is, but how well it’s performing?  And so the two examples that Katie mentioned, looking at power and stability, that’s – how is that electricity infrastructure performing?  When we look at road quality it’s understanding, well, how are roads changing with seasonal weather patterns?  How are some roads that are of some materials versus other materials, how are those performing differently?  But we want to try to fill in some of the data gaps.  We don’t necessarily have sensors in the roadways.  We don’t have tons of vehicles with smartphones that are traveling all of these intercity roadways on the African continent.  We’re not collecting that data necessarily.  But can we fill in part of that picture?  Can we get 80% of the way by using satellite data in a way that hasn’t been used in industrialized settings?  We really look for those opportunities where data can get us most of the way without having to put in costly sensors or other equipment to measure how well that infrastructure has performed.


KATIE:  Yeah, that makes a lot of sense.  And, I mean, related to that, another area that you and your team have dug into which is super interesting to me is demand-stimulation.  So for context, as countries kind of build out electricity grids, especially to more rural and maybe lower-income areas, there’s often very little actual electricity being consumed, sometimes none at all, by these newly-connected households, which creates all sorts of problems for the utility and really doesn’t do very much for development.  And I think everyone in this community is acknowledging that this is an issue, but you’re actually focused on solutions.  So I’m curious kind of what are the most promising ways to address demand and get it up, and how are you thinking about this problem?


JAY:  Demand-stimulation is an area that is crucial not only to the success and growth of livelihoods in the communities, but also for the providers of electricity.  It is important for electricity providers–be they utilities that are state-run or private, be they (biker) utilities that are operating small microgrids, be they small solar home system vendors that are selling individual units–for those companies to survive, and thrive, and grow, and be able to make better product offerings and so on, they need customers who can pay, they need industries that get built up.  And so the other side of the supply coin, whether it’s putting up big generation facilities and the transition to get that electricity out, or whether it’s getting these distributed energy systems, in all of these circumstances you need customers who are growing and using more energy.  And while energy efficiency is important and so on, that’s not the challenge in a lot of electricity-access communities.  It’s really about growing useful electricity that people can then put towards economic purposes, improve their lives and livelihood…  And so as far as how this looks in growth of electricity I kind of separate it into two categories.  One is organic growth, really growing economies.  And we can imagine now that I can process more grain I can start this new business in warehousing, or agri-processing, or whatever else – I could start a new business and kind of grow the pie.  Then the other is conversion, thinking about as we migrate the economy more towards electricity as an energy supply and away from fossil fuels.  Then it’s thinking about, okay, can we convert transportation to electricity?  Can we convert diesel-based milling and agri-processing to electricity?  And so the energy pie might actually be a little bit smaller then because the efficiency of electric motors is higher than the efficiency of diesel-based motors.  But moving it to electricity systems rather than fossil, because fossil tends to be imported (inaudible) energy and has all of those attached costs to it, whereas electricity is almost always locally-generated.  You’re not bringing in electricity from a continent away; you’re bringing it from that area.  And so it’s – both of these contribute, both the organic growth and the converted growth contribute to the health of electricity systems and can contribute to the local economies.  So we really need to focus on not just growing supply, but also having uses of that electricity once it’s in place.


KATIE:  Yeah.  I mean, I think what’s challenging about this is it’s a bit of a sequencing chicken-and-egg problem, right?  So how do you think about growing demand and actually connecting people for the first time, and doing that in tandem?


JAY:  I think we are horribly behind on the demand side.  And so the challenge is that if you look at demand projects coming from the IEA, or coming from utilities, or whomever else, they are almost always rosy.  And the reason they’re rosy is it looks better to investors, it looks better to planners when you have that line going up and to the right.  But that’s not natural.  People aren’t necessarily going to consume electricity if it’s there.  There’s a whole process around this.  And I think we often fail to recognize that this process happened elsewhere, too.  There were huge campaigns in the U.S. (and) the (rural) electrification era to grow consumption via appliances, via electrification of industry, things like that, that those campaigns need to happen but they need to happen differently given how economies have changed, given how rural and peri-urban economies, say, in Africa look differently from how rural and peri-urban and farm economies looked in the U.S. circa 1950.  So I think there needs to be more appreciation that we’ve got different challenges in how we grow electricity demand in modern times, but that it’s not just an “if you build it they will consume” situation.  It’s crucial that we actively push for this, and that active push has to be throughout.  It has to be coordinated government efforts.  It has to be the utility – understanding that they’re not just an infrastructure provider, but they really need to contribute to economic growth.  And the micro-utilities as well.  They can’t end at “we brought the wires and that’s enough.”


ROSE:  Really well-put.  Since solving economic development is such a big problem that is way beyond the (inaudible) of just us power people I’ll maybe bring the conversation back a little bit to something more solidly in our wheelhouse.  So you’re working to harness mobile phones and other low-cost tech to help utilities monitor their grids without expensive smart meters.  Can you share your favorite example from this body of work and why you’re excited about it?


JAY:  So we had worked previously with a system called “grid-watch.”  So a grid-watch was a way to passively use sensors from smartphones.  And we have tons of tech coming up at the consumer edges of systems.  You have really, really cool phones that are cheap now, and everybody wants to have them because they are not just a phone but also a way to connect with the internet.  (It’s got) all kinds of communication and everything else.  But there’s also tons of sensors in those devices.  And so if we’re able to leverage that information – this is another one of those “can we do 80% as well with getting data from phones as we did from having sensors embedded in the network?”  And so the way grid-watch worked was essentially it used all of the passive sensors that are in smartphones–things like the Wi-Fi sensor to see if there’s Wi-Fi networks available, things like light sensors to see if there’s light in the room–and when we saw a coordinated change across those sensors–maybe the Wi-Fi network population changed quickly, maybe the lights also went out at the same time–we actually were able to, with pretty high accuracy, hey, this is an electricity outage.  We’re seeing an outage (of that).  We should note that down.  And when you collect that at lots of different places and you can see if multiple phones see the same event at the same time, that’s not just somebody running out of credit on their prepaid electricity meter, that’s – there’s a system-wide problem and it actually covers the scope.  Maybe it’s one circuit on a (transformer), maybe it’s a whole transformer going out at the same time.  But you look for those coordinated changes across sensors–it’s a way to kind of augment the utility sensing with end-user-based sensing that allows for monitoring of the electricity system.


KATIE:  So Jay, we’ve talked about kind of using cell phone data.  We’ve talked a little bit about satellite imagery.  I’m curious, as you look forward over the next ten years, is there one particular either data analytics tool, or a new emerging method that you think is going to be the biggest game-changer in your space?


JAY:  It’s a great question.  I will say the satellite data that’s evolution in all of the things that we measure is going to change how we look at infrastructure.  I think we’re very much at the tip of what’s a larger iceberg.  We’re getting higher-resolution data every year, we’re measuring more often so the revisit time of these satellites is growing.  You can now buy relatively cheaply imagery that is generated every single day.  You can start to pair together…  So the real value comes not from an individual sensor, but when you start to pair together sensors or images across different domains and draw insights from those…  So we have some recent work–we’re actually pairing together imagery data and pollution data to try and locate irrigation–we’re working in Ethiopia–where irrigation is happening.  But not just that there’s irrigation there, we’re looking–that’s seasonal, like, can you see when there’s dry season in agriculture–but also is that irrigation powered by diesel?  Because we can look at the pollution data to see, okay, that’s actually diesel-based irrigation.  Knowing where that’s happening means you have economies that somebody’s willing to buy that diesel to irrigate there, but then that’s a great opportunity to do that conversion move from diesel to an electric system.  And so it’s pairing together these sorts of things.  This is the real landscape that we’re going to get to grow into is when we start to pull together data sources from all over.  And so that’s where I think it’s – while the machine learning and stuff is super cool, it’s really about being creative and understanding the domain where there’s opportunities to get insights in the domain where it’s about grabbing data from all kinds of different sources.




ROSE:  Coming up, we ask Jay how he’s become the rarest of creatures: an engineer in the development world.  And of course we play Rant or Rave.




ROSE:  So Jay, you have a very unusual profile for an engineer.  You address human well-being questions that are usually posed by social scientists.  You move really effectively within the development community.  And you have a lot of partnerships in Africa, all while pursuing a rigorous technical research agenda within the ivory tower.  So speaking as an engineer myself I found myself exiled to some kind of no-man’s land between academic research and policy, which is true of many of our peers who are interested in development and social impact.  So how have you managed to strike this unique balance between academic research, development sector, and real-world impact?  Like what is your secret?


JAY:  At core engineering is about problem-solving.  And I think it’s important to recognize that the world has changed.  The younger generations are really keen on solving problems that they see, that they experience, that they learn about.  Information – we’re awash in information–now it’s about organizing that information and making use of it.  So I see myself at the beginning of what will probably take over a lot more of engineering departments.  It used to be we would celebrate the scientists and engineers who worked tirelessly in the lab by themselves and, you know, blocked out the rest of the world.  That’s no longer the way that engineering can work.  We really need to find the problems out there, solve the problems out there, and have context beyond a kind of narrow focus in one particular area, whatever else.  While it’s important to make that (shift), you know, 5% more efficient, 10% more efficient, that’s only part of engineering, not all of engineering.  And so I think there’s lots of energy from government, from the donor community, from private companies to really look at the world’s problems and address them more systematically, more holistically.  And my research aims to – our team’s research aims to not just focus on the technical problems behind how do we make use of these data sets, but really why do we care about that?  What’s important?  What does it improve in the world?  And so it was crucial to understand context just as much as it was to understand the technology and how things are changing, and how to do systems-based research.  And so that’s what we in our group tend to focus on is understanding both the newest techniques of how to work with data sets, but also what are the problems for infrastructure and electricity systems and how are those evolving in different parts of the world?


KATIE:  Yeah.  So related to that, one of the coolest things that I’ve seen about your work is that in addition to pursuing your own research you’ve also aggressively recruited a ton of young African grad students to study with you at UMass Amherst, which is an incredibly prestigious school, and it also I imagine takes a lot of your own time and effort to find and attract those students.  So I’d love to know kind of what are the most important qualities or insights that those students are bringing to your program.  Why have you found it so important to recruit them?


JAY:  Frankly I actually see it reverse – it would be irresponsible to be working on research solely in Africa without African perspectives.  And that is a commitment of me.  I work with a number of other academics in the U.S. who also are committed to the same ideals and we actually have a joint consortium where most of our students are African graduate students.  The world is as moved where  those pathways are being built that there’s a path to – from the top schools on the continent to the top schools in the U.S., and onward.  And so the perspectives you gain, whether it’s understanding the nature of the real challenges–that demand is actually the most crucial problem and supply is the relatively easier one to solve–versus what are the different challenges in rural settings, and Uganda got that versus peri-urban settings, and Uganda versus urban settings, and Uganda…  Understanding that continuum is not something that we can learn growing up in the U.S.  You need to have the varied set of experiences in Uganda to be able to understand that context.  And so it would be – it would take important dimensions away from the kind of work that our team does if it weren’t being done by Africans.  So it’s much more of a double-negative there rather than a…  Well, we’ve got to have more Africans that are pursuing this kind fo research, because the creativity comes from understanding the context as well as the technical pieces that we put together.


KATIE:  Absolutely.


ROSE:  Okay, so clearly you have a mission to use your technical skills to address big problems in Africa and developing countries.  Where did this passion come from?  I don’t know, is this – was this taught in Arizona public school?  Like, where and how?


JAY:  Yeah, so growing up my parents are from India, they moved to the U.S. in the ’70s.  But most of my extended family is in India.  And I visited as a kid.  I was witness to an incredible groundswell of economic developments.  I have, you know, images as a kid of driving by–my family was primarily in New Delhi, so big megacity–driving by slums that were poor.  And our family is very well off, we have tons of – very lucky to have that privilege.  But to see the range of – it’s said that you can see kind of ten centuries of development within a square kilometer in India, just because you have high-rises right next to slums, right next to informal settlements.  And to understand how that evolution has happened, India’s not like that anymore.  India has – while there’s still tons of poverty in India there’s a ton of wealth and it’s locally-generated wealth.  And so to see that evolution helped me get a better appreciation of how open economies and technology development can really improve livelihoods, while the pandemic has been an awful process.  The hope is that it’s a blip and we actually can continue what’s been an incredible growth for billions of people across the planet to move into more comfortable, stable, higher-well-being lives.  And so this I think is an opportunity to channel some of that effort towards the most salient problems when it comes to electricity and – not just for well-being and development, but also for clean electricity that can enable those economies in a way that doesn’t damage the biosphere.


KATIE:  So one last kind of big-picture question.  You’ve done so much interesting work on a variety of topics.  What are some of the biggest – maybe the one biggest unanswered question that you want to investigate in the next decade or so?


JAY:  Great question; hard question.  I’d say where we still don’t understand demand is why isn’t it growing?  We have every amount of examples, we just haven’t been able to sort out what combination of infrastructure systems allows economies to grow.  I don’t think we can – even as we’re energy people primarily, I don’t think we can run away from the economic development question.  And I’m not trying to say we are running away, but I think that if we aren’t pursuing that question then we are missing the forest for the trees.  What is the combination of investments in infrastructure–both hard physical systems, virtual systems, data, things like that–and education and so on for people to make use of those systems that actually unlocks growth.  What can we do to best have that happen?  We still haven’t solved this problem.  We’ve seen fits and starts in development, we’ve seen places that have excellent roads but poor human capital, and maybe poor energy systems, or maybe places that have great agricultural infrastructure and markets and so on, but really poor electricity systems.  What is the mix of infrastructure systems that really unlocks growth and allows people to achieve those benefits both in economic terms as well as well-being, and happiness, and comfort, and everything that comes with that?  But we still haven’t sorted that out.  I think that’s a guiding question for careers, not just for, you know, five, ten years.  This is how do we think about the mix of infrastructure and development in a way that can lead towards all of the benefits that come from a high-energy, high economic growth, and well-developed – great human development kind of society.


KATIE:  Huge question, and I’m glad you’re on it.  Thanks Jay.




KATIE:  Okay Jay, now it’s time to play Rant or Rave, a quickfire game where we give you a word or a term and you choose to either rant or rave about it, and explain why.  So we’ll give you a minute and thirty seconds and this is all about first reactions.  We’ll get through as many as we can.  Are you ready?


JAY:  On the clock, let’s do it.


KATIE:  All right?  Ready, go.  Undergrounding electricity distribution?


JAY:  Rave.  We can get fewer outages and better performance out of the system.


ROSE:  Electric vehicles.


JAY:  Rave.  Decarbonizing transportation, hugely important.  We don’t have any other better routes that can do this.  So it needs to grow massively.  It’s already happening in Norway, take a look.


KATIE:  New England winters.


JAY:  Rant.  My snot froze this morning, but opportunity to really look at how we change the heating systems in homes.  We can really push heat pumps and things like that.  So there’s lots of opportunities for energy professionals to kind of improve those winters.


KATIE:  Good job tying it back to energy.


ROSE:  Making IEA’s data freely available.


JAY:  I’ll say it’s – the idea’s a rave.  I’ll go with the rant of we need data to flow, especially massive data like that.  And if they don’t make it available then I will come up with other ways to come up with better data.


KATIE:  Okay.  Nairobi traffic.


JAY:  Rant.  I was caught in traffic for six hours one time because of the flood.  It was – and it was – actually it was eight hours.  It was 6:00pm I left work, I got home at 2:00am.  It’s single-tracked.  You’re on two lanes.  It’s this mix of being polite and not honking at people, but passing on the left, passing on the right.  It’s not necessarily an infrastructure problem as much as it’s how people drive that causes the issues.


KATIE:  All right, we’re out of time.  That was awesome, Jay, great job.


ROSE:  Jay, thanks so much for being with us on High Energy Planet.  We really enjoyed this discussion and we look forward to bringing you back on in the next let’s say one year, and you can tell us how you solved demand (inaudible).


JAY:  Honored to be here.  Thank you very much for the opportunity.


KATIE:  Bye Jay.




KATIE:  That’s it for today’s show.  High Energy Planet is a production of the Energy for Growth Hub, matching policymakers with evidence-based pathways to a high-energy future for everyone.  Find out more at and tweet your questions and thoughts to us at @EnergyforGrowth.


ROSE:  If you liked today’s episode be sure to rate and rank the podcast and tell a friend about us.  Bob Lalasz is our executive producer.  Mariel Ferragamo is our senior producer.  Join us next time for more High Energy Planet.