Rajat Monga is an Electrical Engineer from IIT Delhi (duh), batch of ‘96. From the halls of Karakoram, he’s now an entrepreneur based out of the US. He’s worked with some of the firms that are leaders in technology and its applications, like search at EBay, or Risk Management at Morgan Stanley. He’s been the head of TensorFlow (which he co-founded) at Google - the machine learning platform used worldwide. He’s also been involved in building startups from the ground up, something he’s doing currently in the form of a stealth startup.
Contrast your expectations at the beginning, while joining IITD vs at the end, what do you think the biggest differences were?
The big expectation that wasn’t met?
Coming in, I thought I’d learn a lot!
I mean, I did learn, but in very different ways. I think the blame for that falls on both ends. The college courses weren’t as exciting, but I didn’t invest the time or effort either. In most things, there’s a fun element, and some effort you need to put in to get to that fun – and I wasn’t willing to put in that effort at that stage in my life.
On the positive though, I did have a lot of fun, made a lot of friends, many of whom I’m still in touch with. The people I’ve met, and all the things we did – like DumC’s (Dumb Charades), or various types of puzzles. The fun we had and the things we learnt made it a great experience, really.
On that note, what places and incidents do you associate with fun at IITD?
A couple of things. First off, we would bike everywhere, to movies at CP late at night. It’d take like an hour to get there – but it was great fun in large groups.
During late nights and stuff, the parantha place outside the gate, under the flyover (Sassi da Dhaba) – that was a great place to be. We’d go there often for anda paranthas and chai – especially when there were minors or majors. It’d be 1 am, and we’d just started studying because we were doing other stuff all day, when someone would say “kuch toh khaate hai jaake”. Then we’d eat and go to bed because it was pretty late anyway :P.
The elephant in the room – could you tell us about your career path into ML, which led to heading TensorFlow, starting from EE at IITD?
While I was from EE, I was always very interested in CS. I even took a couple of CS courses, although I wouldn’t say that mattered too much in the larger scheme of things.
At the time, for EE, companies like Cadence and others were the top choice. Those weren’t super interesting at the time; my top choices were companies like Infosys or TCS. So, I ended up in Infosys, and to be honest the job wasn’t as exciting. I did some fun stuff, worked on the Y2K problem, learnt a few things. Eventually though, I ran out my one-year lock in period and joined a small start-up in Bangalore. I learnt a lot about working with people here, which is probably as important as the tech aspect of it. Eventually, through work, I ended up in the US, and I’ve been here ever since.
I landed up at Google 10-11 years ago, in the ads group, where I learnt a lot simply from the scale of operation. Couple years in, I wanted to do something new, looked around within google, and found this completely new space – deep learning. In hindsight, I’ve taken a lot of bets, but this one worked out well!
Eventually, that developed into TensorFlow, where I worked a bit on the software side initially. As time passed, my focus did start to become more about where we wanted the product to go and planning rather than just software.
Towards the end of 2019, I decided to work on my own start-up, and now I’m in the process of doing that.
What’s your opinion on the rush toward ML for most of the IIT crowd today?
I think there’s a problem with the way we bucket people at IITs. There’s a fixed number of seats for CS, EE, Chemical and so on. It's not like you don’t want people to learn these things, things I learnt in EE did help me when working with hardware teams at google, and stuff like. But this also means people are constrained in some sense.
The one thing I will say about ML is that you’re layering things on top of each other. So it depends on the field you’re trying to get into – if you’re applying ML, you need to be familiar with the software side of things. If you’re looking at the statistical end – you need to know the math behind it. It gets easy nowadays with things just being packaged together, but if you want to be successful in the long run you need to be able to think through all the layers and have that background knowledge.
Is ML that important to learn, though?
I would say yes. It’s getting to a point where everything is going to be driven by software, so understanding that end of things, even if it is not your focus, is important. Even if you’re a physics major and want to stay in that field forever, understanding software and possibly being able to program is going to help you. For example, it's moving on from just basic simulations and stuff, to applications in weather prediction, or fluid dynamics, where people are starting to apply ML and see exciting results.
What’s your opinion on creating a start-up culture in the IITs? Why do you think, despite efforts to create it, it doesn’t seem to be moving as fast in India?
In my opinion, it’s a bit unfair to expect undergrads to take the dive right after a B.Tech and start a company. Not that they shouldn’t, it’s a good push and it takes time. Comparing it to the US, for example, Berkeley and Stanford see a lot of dropouts who start something. However, the larger percentage is the masters, or PhD students who do that.
That’s generally because they have the financial ability to do it, and they have a bit more experience. Compare that to an IIT, where your parents have probably paid for your education, and if you start something – that’s asking your parents to pay for a couple years more. It’s not something that the average kid necessarily wants.
What I would love to see is IIT encouraging a lot more of the graduate stuff. Other than that, innovation from a primarily technical institute should be technical. While it does happen, it's rarer to see a purely business model innovation. The best place to get that technical innovation is going to then be graduates. It’d be great for undergraduates to also be involved.The other thing that’s nice is stuff like axlr8r.
There’s an extreme view where you just give students money to drop out and do their things, but it’s not one I buy into yet.
On a similar note, do you think there is a lack of interest in technical education in IITs today, has it changed from your time?
I don’t know much about the landscape today, but I will say that it was not great even in my time. Again, I think the problem comes from the IITs bucketing people, in some sense, because it is hard to tell someone to remain on the technical side of things when they end up in textile engineering. So, in that sense, I would love to see the institute encourage people to remain interested in technology.
In my experience, often people who didn’t go to an IIT, who could be dismissed unfairly as “not as smart”, become successful in their tech fields because of the interest they have and the effort they put in. I think it's fair to expect that there should be a proportion of the IIT populace that also has that sort of interest, which is where I think the system needs to improve. At a personal level too, I think a lot of IIT kids who are interested in tech fields would be happier staying in their fields rather than chasing MBAs, or placements, for example.
Do you have any “fundae” for the general IIT populace today?
Wow, that’s a very general question! I could go on and on, probably.
One of my regrets at IITD was probably just coasting a little too much, where I did not take enough advantage of the facilities around me. It's easy to blame the college in terms of a lack of infrastructure, or a poor system – but eventually we do need to take responsibility for what we do.
Other than that, I think it is important for IITD students to realise that learning doesn’t end here, at the end of 4 years. You haven’t started a career yet; this is just baby steps into the world. We talked about Machine Learning, and yes, that’s the big thing today, but it could be something else in a few years, which means that being able to continuously learn is an important skill to have.
Last question, what is the delta (positive or negative) of IITD in your life?
I’ve never really thought of it that way, but on first thought I’d probably say the friends I made there. There were other things, a few positive learnings for me, but nothing negative really.
Initially, we couldn’t stay in touch as much. Now, with things like WhatsApp, we’re able to stay in touch, a number of folks meet up every year, and I’m able to join them every once in a while. Those connections, those friends I made are probably the biggest positive delta by far.