Bangalore Startups’ AMA with Ashwini Asokan & Anand Chandrasekaran, Founders of Mad Street Den
Mad Street Den, a one-year-old artificial intelligence startup based in Chennai, started by husband and wife duo who returned to India from Silicon Valley last year has closed a $1.5 million seed round to bring technology developed in the labs to consumers in a useful way.
Anand is a Neuroscientist turned Neuromorphic Engineer from Stanford, who was a part of the team that built Neurogrid. Ashwini is a designer turned Mobile Innovation Lead at Intel Labs, driving cutting edge mobile research and working with a team of engineers, scientists, designers and social scientists to bring machine learning, sensing, image processing systems to life in the form of consumer centric products.
MSD offers AI and Computer Vision as a plug and play solution to businesses and their platform — the MADStack offers a wide range of computer vision modules(Gesture & Expression detection, Visual Search, Emotion detection & Gaze tracking to name a few) to deliver exciting new experiences for their consumers. Just plug into their AI cloud platform with an API and you’re all set!
Q: Welcome Ashwini & Anand to the AMA, tell us about your “MAD” story. How did it happen? — Mithun Muddan
We were going through a classic phase in our lives. Anand, at the end of his academic career and me coming up on 10 years at Intel. We’d been talking about the future of AI — him from the tech perspective and me from the people / products perspective and one day we said, let’s just go do this ourselves and packed up 15-ish yrs in the US and just left. So we came back with the aim of building an end to end computer vision platform which is where we also will integrate the AI we build.
Q: Where should one start if one wants to work on deep learning projects? — Madhav Kandukur
Deep learning is easy to get started with, there are numerous tutorials and tools to get going, try Caffe, Theano etc. The tutorials will get you started.
Q: Can ya give us an insight into your platform’s technology stack? — Karthik Nagarajan
Our platform, the MADstack currently offers several computer vision modules, with some familiar ones like visual search, gesture recognition, expressions etc. The goal is to provide these and more in a coherent single platform that delivers all of these in real time.
Q: Can we see a demo of your recognition technology? — Richa Kanna
Unfortunately our APIs are open only to the customers we’re working with right now.
Q: How are you getting along with problems like industry adaptation for example fashion as part of your product line? — Sreechand Tavva
Yes working in each geography has it’s own challenges. And these challenges are especially hard when you’re trying to create a category in the market. While a lot of these technologies have been around for a while & AI itself as a whole, if you think about it — the use cases or applications and the method of implementation is just as important as the technology itself. We’ve been trying to solve a lot of these challenges by working very closely with our business partners.
Q: How are you different from Google Prediction API and Amazon’s new ML API? — Kartik Mandaville
As opposed to those platforms which aim to provide a general framework that could be applied to a lot of different solutions if you know what you are doing, we are building solutions that solve specific problems that we think the world is ready for.
Q: What were the initial challenges that you guys faced or still is an issue in your growth? — Chiranjeevee
Well we were coming back after being away all our adult life, so to begin with I think the move back itself was a huge challenge. I think we often underestimate the role of geography and culture and this is one of those lessons that we learnt the hard way. We took almost a year to adjust and get running. Thankfully our experiments worked very quickly and we hit the ground running pretty soon after. Second challenge was in figuring out what use cases worked and didn’t work. This is one of those things that I think we’re continuing to explore and learn. Technology alone will not cut it, although we believe we’re here to build tech that beats competition of course. But we’re focusing pretty heavily on understanding and implementing tech in a meaningful way. This often means, failing over and over again before figuring out product-market fit. We did that very rapidly in the first year.
Q: Why India? Isn’t US / Europe an easy market for these services? Do you have Indian customers? — Vignesh
We believe that in India, there is the potential to leapfrog adoption of new technology and with the rapid growth of consumer tech this place is ripe for category creation.
Q: Since there is a lot of research already happening in US for AI, how do you guys plan to keep yourself updated with latest in thing in tech? Does India have a right competencies to take it to the next level? — Sohil
Well we’ve spent a lot of time in the US in these fields. Anand is well connected in his circles and me, through the companies that I’ve worked with in some way in the last 10 years. We’re also constantly doing our homework, understanding who’s who, what’s going on in the market. Investor community, customers are another great source to understand what’s going on in the US market. That said, you never know — everyone seems to be in stealth mode these days.
Q: I’ve studied a bit of Machine Learning in the past and I’m interested to see how it proceeds when it comes to Experience Design. How do you conduct User Validation for the new features that you launch, and is there a constant procedure you guys follow? Also, how difficult can it be to actually implement the results obtained from the User Validation into the AI? — Swapnil Borkar
I’d love to chat with you about this offline if interested. This is a philosophy we breathe at Mad Street Den. We believe that tech without people is meaningless. And we constantly strive to make the tech we build useful and exciting for people / customers. And of course we have an in-house design and UX team that constantly puts thinking about people and nature of use of technology, much much ahead of the actual development.
Q: What are some examples of gaming use-cases for your current tech?(I run a gaming startup, hence the interest) — Gautham Raja
We’d love to chat about this offline. But there are several use cases that involve using gestures, expressions making fun/social games. Do ping us if interested.
Q: Are Google’s and IBM’s machine learning efforts going to be direct competitors? — Madhav Kandukuri
Some of the most important work in AI and CV is being done at Google, IBM, Facebook etc. We are well aware of what is going on, but the space is vast, and just being entered. All of them, and us, have barely scratched the surface.
Q: None of the predictions made decades ago in science fiction are anywhere near coming true. What is your view on the other side of artificial intelligence, the useful side we could have in everyday life? — Karthik Nagarajan
Yes this is interesting isn’t it?! Lots of sociologists and anthropologists have written about this topic. You should go and see if you can get your hands on a paper called Yesterday’s tomorrows by Genevieve Bell and Paul Dourish published at Ubicomp. Lots of lessons to learn on how a tech field like AI can fail over and over again if not cast in the right context. Google glasses are a great example of how a tech can fail sometimes even if cool and futuristic. Walking around wearing something like that and ID-ing everyone and everything is far from the world we live in as humans. We can’t sit in a hole somewhere and dream up the future of tech. It has to be embedded within the context of stories and life of people.
Q: How are you handling hiring? How do you find people capable of working on the problems that you are working on? — Kunal Jain
Its never easy, and it shouldn’t be. Having said that, when we present our story to the world, like minded people seek us out more often than not. It also helps to have your well wishers search for you (investors etc.) Also, our journey in India has been very interesting. Being in Chennai has surprisingly helped us hire faster than when we were in Bangalore. We believe that competing in the Bangalore hiring space with ‘rockstar unicorns’ makes for difficult hiring.
Q: Given the background in neurosciences what do you make of Sparsed distributed representation techniques by Jeff Hawkins and even www.cortical.io team? — Keshav Meda
Without going into specifics, I draw a lot of inspiration from Jeff Hawkins, as do a lot of others.
Q: Who are your competitors in this space? I haven’t seen another start up trying to do exactly what you guys are going for! — Karthik Nagarajan
We have lots of competitors. On one hand, we have startups competing on each vertical we’re into. So Cortexica, Camfind are examples on the visual search space. Affectiva in the emotion / expression detection space and several more as you can imagine. On the other hand, we keep coming back to the fact that this is a vast space and by no means, we’ve even started hitting the basic use cases for the market.
Q: What was the inspiration behind the caret logo? It rings a bell, because a similar company in India uses the same and I’m interested to know your story behind it. — Swapnil Borkar
That was coincidence, we have been exploring various logo options for months, and we are not sure we are set.
Q: What are the applications of your platform targeting automobiles? — Rehan Asif
We’re not hitting auto just yet. I think that’s one we’re probably going to wait out for a bit. I’m not sure the market is ready to bite on that one.
Q: Could you highlight some positives and negatives about having your partner as cofounder and working on a startup? What have you guys learnt so far that you didn’t think of before starting up? — Prabhu
Well for starters I would say, don’t do this with your spouse if you haven’t spent adequate time with them before the company. Also, don’t do this if you don’t share the same passion / vision. It’s not easy. We’ve got 2 kids — a 5 yr old and a 9 month old and you can imagine that life is just intense on and off work. It’s not easy but we’d have it no other way. It’s hard enough trying to make marriage work, I’m not sure why anyone would find another co-founder who they’re not married to:-P
We’re pretty much attached at the hip both at home and work☺
Q: How is the adoption rate of AI systems in India ? — Bhuvanesh
The market for AI is just opening up. We’re barely scratching the surface, anywhere in the world. The basic big data based based analytics is definitely much more advanced than any other kind of AI. And it’s still very basic. The good thing about India specifically though, is that the leapfrogging is really helping. Companies have taken to our offerings a lot quicker than we ever imagined. I think it’s going to be a quick, steep adoption curve
Q: How important is it for your business that your algorithms are the best in the world at what they are doing? How does your system do against the current best benchmarks? — Madhav Kandukuri
For us, on one hand, our PoCs prove their worth to our customers, instead of numbers. Numbers are abstract things. And though lots of people think benchmarks are the way to go, we decided to let our applications do the talking. And saw a huge difference. Most of our businesses have actually tried almost all of our competition from across the globe, and we’ve heard some good feedback and stood out. Going out and experimenting in the market is all that matters in new categories like these. You’ll have some successes, some failures, you’ll learn and move on. It’s really too soon to say what works and what doesn’t. That’s the mindset with which we’re approaching this. Yes we can benchmark ourselves and tell you how we’re doing with Image Net but that’s not our strategy / approach to AI. Image Net is but, one tiny piece of the AI approach. We all tend to get too hung up on this and need to approach AI more holistically beyond just this.
Q: How is the neuroscience background useful in building machine learning/computer vision tools? Have you found any good concrete relationship between modern AI and Neuroscience? — Navneet Sharma
The Neuroscience community has known for decades that even in early stages of visual processing there are ten times more feedback connections from the brain than the feed forward connections from sensory inputs. If you think of several techniques in modern ML including deep learning, they have a purely a feed forward architecture which negates the possibility of including things like context or attention modulation. These would have very powerful implications and immediate performance boosts if incorporated the right way in any neural architecture. As a result, I’m one of those people that think there is a ton of knowledge from Neuroscience & a lot of techniques from ML that can work complimentarily.
Q: When it comes to AI, it did well with gaming, analytics, IoT and social media, because there is lot of data to analyse and interpret while it comes to fashion there is not so much data but more of need at least when you look at the Indian e-comm market, where you can identify a product just by looking at, How’re you addressing this case? — Sreechand Tavva
All of this really does go back to my previous point on experimenting. Yes there aren’t complete metrics in the e-com market specifically. But there are some. And if your PoCs really are worth it, your customers will share them with you, whatever they can. It is in their interest to allow for the kind of personalization that AI can offer. We’re literally doing this — one customer at a time. We’re working closely with each and everyone one of them in our early days now. There are no easy answers here. We’re not yet in an environment where these things are established and we can all run off of business models and product approaches that have been established. The people in this game now, ARE the ones setting the metrics, the models, the use cases.
Q: Is India your target market or are you guys also trying to sell your product in a hyper-competitive US market? —Navneet Sharma
We’re a global company. We already have team members in the US despite having our main presence here in Chennai. We’re from there, our connections, networks are all there. We’ll be there, here, in Europe, Asia or wherever else there is demand … that’s the hope. But we see a really exciting time here in India and we’re coming across some wonderful opportunities as our experiments are turning into actual products that are being deployed.
Q: You guys got patents and do you think that’s necessary for AI startups like yours to build competitive advantage? — Richa Kanna
It’s necessary for any product startup to be built on sound / defensible tech these days. Patents or no patents, if you don’t have something that others will find difficult to catch up to if your focus is building tech or tech platforms, then there’s really no hope these days. In a world where Whatsapp and WeChat are disrupting e-commerce and Uber & their likes, where social media companies like Facebook are turning into treasure trove for AI .. there’s no telling where you’ll be disrupted from. Sound / defensible tech goes a long long way in giving you a chance for survival.
Q: If you guys are hiring do let everyone know. I’m sure there are talented people here. — Kartik Luke Singh
We are always hiring, and across the board. Backend, Frontend, Data, CV & ML. For those of you guys that have asked for email ID for getting in touch — info@madstreetden.com would be a good place to start. Do just reference the slack conversation there so we can get back.
That brought our AMA with Ashwini & Anand to a close!
Final words from Ashwini & Anand: Thanks Guys, that was fun and thanks for all the wishes☺
We want to thank Ashwini & Anand for taking the time to chat with us!
We’ve set up a Twitter handle — @blr_startups. We’ll be tweeting updates and such out there.
If you are interested in the startup ecosystem and would like to network with fellow tech founders join the Slack group if you haven’t already! Just fill out this form. If you have already filled the form, you would have received an invite. Contact @kar2905 or @harshamv if you have any issues. This AMA was summarised by @kartikluke and @karthik2502.
You can suggest other people for AMAs at the Slack group or you could reach out on Twitter.
