Home / Technology / Learning AI from the best profs just got easier

For Sakthisree Venkatesan, the covid pandemic arrived just when she had taken up a machine learning job with a wholesale retail chain after graduating in computer science from an engineering college in Coimbatore. It meant she had to work from home, and saved time on commutes to the office, which she put to good use by signing up for an artificial intelligence course on It’s a newly launched platform where professors from the world’s top educational institutions teach AI from the foundational to advanced levels.

For Venkatesan, this was a different experience from the MOOCs (massive open online courses) she had signed up for earlier, including Stanford professor Andrew Ng’s famous machine learning course on Coursera. “Unlike pre-recorded lectures, lets me interact with the professors. We have live lectures and Q&A sessions where we can ask them about the latest research in AI and not just the coursework. A lot of my learning also comes from my peers and the projects we do together," says Venkatesan.

Her first teacher at was Rahul Dave, who was in the original team of Harvard’s CS109 data science course. Dave left Harvard in 2019 to co-found with Siddharth Das, an MIT alumnus who was a co-CEO of Reliance Jio Money and a COO at Flipkart before that. “A climbing buddy introduced me to Siddharth who already had this idea floating around in his head from seeing the students coming to work at Jio. I thought this would be an exciting thing to do in India where we have a large talent pool of smart kids who don’t have access to high quality education," says Dave.

The founding faculty at includes Pavlos Protopapas, who leads the data science masters programme at Harvard, Achuta Kadambi, who leads the visual machines group at UCLA, and Raghu Meka, associate professor of computer science at UCLA. Biocon founder and chairperson Kiran Mazumdar Shaw is its first angel investor. A larger fundraise involving institutional investors is currently underway.

“I settled on AI not because there’s so much hype around it but because fundamentally there’s a paradigm shift," says Das. “At the simplest level, instead of writing a set of rules for a computer, we give it an objective and let it optimize a programme to meet that objective."

Super chips combined with AI modeling is taking our ability to extract insights from massive amounts of data to new levels. Yet, “whatever we see today, minus the hype is just a start because we’ve found only one or two paradigms to simulate how humans probably think," points out Das.

In such a scenario, where research and commercial deployment are both evolving fast, it’s not surprising there’s a dearth of quality in faculty and curricula. “Largely worldwide, and most definitely in India, the pedagogical teaching resources for this new science just do not exist," says Das.

Apart from MOOCs, online platforms like upGrad, Great Learning and Springboard have come up to address gaps in professional education in areas such as data science.’s differentiation at the outset is its drive to build a curriculum and programme with a network of world class professors who are directly engaged. This differs from edtech platforms that essentially connect demand (students and professionals) with supply (vendors and universities offering online versions of their courses).

Such an approach aims to form deeper connections between students and teachers, which is vital in a fast-evolving field like AI/ML where new research quickly outdates curricula. “Each one of our courses provides 8-10 hours of live learning and mentorship per week," points out Das.

Initially, the idea was to conduct bootcamps in Bengaluru, but covid accelerated the move to online. Apart from making the live classes and interactivity seamless for students and teachers logging in from multiple locations, is modularizing its courses to make it easier for the faculty to give their time.

“If you gather professors from around the world, an interesting approach is to design courses such that multiple faculty can teach small slices of a course. Yet, the integrity of the course must be held together," says Das.

A substantial part of the learning comes from project work which reflects what students are likely to come across in real-life jobs. has a consultancy arm that helps corporations from multiple fields apply AI/ML, which in turn helps faculty design homework assignments that are more realistic than the usual textbook or coursework projects.

It’s something Dave has been used to doing from his days at Harvard, where faculty are allowed to take up teaching and consultancy gigs outside., for example, is working on an autoencoder model for an insurance company, which is a form of unsupervised machine learning that automatically removes ‘noise’ from data inputs so that patterns are found efficiently. “I’m already scheming out ways to use this kind of autoencoder for outlier detection in other publicly available datasets for our homework," says Dave.

In fleshing out projects as well as anchoring a course with modules taught by different professors, teaching assistants (TAs) play a pivotal role. Apart from guiding students, they free up time for professors to engage in Q&A sessions, ‘office hours’ mentorship, and bringing new material into courses.

One of the TAs is Anusha Sheth, who was earlier a student in the first cohort of when it was a pilot project conducted in offline mode as a summer camp. She had joined a multinational company as a software developer after graduating in electrical engineering from a college in Bengaluru. “ helped me get back on track in what I actually want to pursue, which is to apply AI to large problems in energy," says Sheth.

Becoming a TA has proved transformative in her own learning path as well. Sheth recalls a backpropagation algorithm on which Dave had asked her to build code. “He gave me so much feedback on the code, the mathematics behind it, how it can be applied to multiple fields, and what needs to be tweaked to get better output. He gave me resources to read and I revised my code. Now if he asks me about backpropagation even in my sleep, I would be able to tell him about the maths and everything about it!"

Sheth will be going to Delft University in the Netherlands for a master’s programme in sustainable energy later this year. thus prepares students for research apart from AI/ML jobs. Venkatesan, for example, has become a research mentee for Pavlos Protopapas at Harvard. She knows it will be a feather in her cap when she pursues higher studies. “I got the opportunity to become a part of Pavlos’ research group of students because I was able to show my skills in the course that he took at"’s first regular batch started just half a year back. The challenge going forward will be to scale up its network of professors and get them to participate in the current involved fashion, given the fact that they would be primarily teaching in their parent institutions.

Das aims to tap the large number of adjunct faculty who work with limited contracts in US universities. “They’re very good but poorly paid. We can hire them and reduce our dependence on star faculty who can be kept for things where their contribution is the highest."

Selecting students with the most aptitude and commitment will also help with peer learning and efficiency. Already, even though it has just got going, there are over 10,000 applications, says Das. That’s not surprising given the current faculty.

For Venkatesan, a high point came recently in a live lecture and Q&A by Geoffrey Hinton, one of three researchers who won the 2018 Turing award. He spoke to the students about his revolutionary new GLOM architecture, which could help simulate how humans visualize an image even if they see only a part of it.

Sumit Chakraberty is a writer based in Bengaluru. Write to him at

Subscribe to Mint Newsletters
* Enter a valid email
* Thank you for subscribing to our newsletter.

Never miss a story! Stay connected and informed with Mint. Download our App Now!!

Edit Profile
My ReadsRedeem a Gift CardLogout