We are Redefining Learning with GenAI and Hyper-Personalisation: Ankur Dhawan, CPTO, upGrad

In the ever-changing online education space, upGrad has been revolutionising professional learning for a digitally native generation. Leading this charge is Ankur Dhawan, Chief Product and Technology Officer (CPTO), upGrad, who is charting the company’s product and tech strategy to combine AI, hyper-personalisation, and learner engagement into a frictionless experience.
Here’s this exclusive interview with Express Computer, with Dhawan discussing how upGrad is employing generative AI for intelligent assessment, developing in-house tools such as CodeX, and achieving an 80%+ course completion rate through nudges, personalisation, and extended learner insights.

How is upGrad using AI and automation to improve the learning experience, simplify operations, and provide personalised education at scale?

We would rather call our users learners or students than customers. AI has caused a tremendous shift in the delivery of education. The transition from offline to online education became especially pronounced during COVID-19, as people realised they could access learning and upskilling opportunities without needing to leave their homes. The arrival of ChatGPT, particularly with its advancements in GPT-4 and generative AI, has further accelerated the potential of personalised learning; something long considered the holy grail of education.
We’ve always asked: can we provide every learner with their own tutor? With generative AI, this has now become far more attainable. We’re using AI, specifically generative AI to answer learners’ queries. For example, a working professional, perhaps married with children, might only find time to study late at night. Earlier, if they had a query, they would have to arrange a call with a teaching assistant, maybe the following day, by which time their enthusiasm may have dissipated. But now, with AI, a bot is at hand to explain a concept in simpler language.

Also, the bot does not rest on providing answers, rather it pulls from multiple sources, even beyond upGrad material, and can even ask follow-up questions to test comprehension. The two-way interaction is either students asking questions and the bot asking them questions about understanding or allows for a tailored experience. The AI learns to know the student and makes adjustments accordingly, thus providing an extremely personalised learning experience.

On the automation front, AI has also enabled 24/7 tech support. If a learner encounters a technical glitch; such as being unable to access a module—the bot can solve and remedy the problem without human involvement. This provides seamless access to learning and re-enforces motivation and engagement.

What is upGrad’s contribution towards enabling professionals to upskill and reskill for new industries, and what are the main challenges and opportunities presented?

As the professional landscape evolves, many existing job roles are becoming obsolete. To remain relevant, individuals must become ten times more effective than they currently are; we call this the ‘10x’ model. Whether it’s a 10x engineer or a 10x marketer, leveraging technological tools like Copilot and others can significantly enhance productivity and professionalism. We are strongly committed to enabling this shift. For example, we have recently partnered with Microsoft to roll out a Copilot certification programme that enables professionals to upskill rapidly in a short period of time and become 10x in their function.

Everything is technology-backed on our side. Upskilling 5 million students, for instance, in generative AI or some other field cannot happen without technology empowerment. As soon as a student comes to our portal, we are obsessed with encouraging him or her with gamification, reminder nudges each day, interactive modules, to finish up an assignment, quizzes, and then a certificate.

We also emphasise applicability: learners must be able to use what they learn in real-world contexts. Our learning management system (LMS), developed in-house, is specifically fine-tuned to align with our courses and delivery models, ensuring a seamless and effective learning journey.

As the CPTO of upGrad, what do you see as the future of online learning, particularly with immersive technologies such as AR and VR? How are you gearing up for this transformation?

Online learning has gone through several stages. In 2020, in the midst of the pandemic, the issue was access at a basic level; many lacked consistent internet or hardware. We facilitated that gap, particularly in K-12, so students could continue their education from home. We gained significant insights around engagement, motivation, and managing distractions.
Now, we discuss immersive technologies such as AR and VR. As promising as these are, so are they tricky. The overwhelming computational requirement for full-fledged AR/VR proves a hardship for most equipment, making learning suffer. So, that’s why we make selective use of AR/VR, largely for gamification.

For instance, we could utilise an avatar to present a complicated subject in a more interesting manner. However, we haven’t found significant learner demand for immersive AR/VR experiences in higher education, especially since many of our programmes are tech-focused and involve coding. For such subjects, understanding development environments and practical coding skills are far more critical than immersive simulations.
That said, AR/VR holds more relevance in K-12 education. For a younger student learning how planets orbit the sun, virtual reality can be a powerful visual tool. But for upGrad’s current offerings and audience, the bandwidth and content suitability of AR/VR remain limited, and hence, we’ve not heavily invested in it yet.

How are you using data analytics and machine learning to create personalised learning paths? Are there any tangible outcomes you can share?

Yes, so I think I’ll come to the outcomes later, but let me just explain what we do.

When a learner joins the upGrad portal and begins their course, we start collecting extensive data like what time of day they usually study, which days they’re most active, how long they spend on a session, which concepts they complete, and their performance on quizzes and assignments. From this, we create a detailed skill report.

This skill report enables students to know their strengths and weaknesses and is very beneficial for him to decide which skills he needs to further strengthen and which ones he is strong at. It is particularly helpful when they are writing their CVs or planning for career changes because it enables them to clearly define particular skills they have acquired through upGrad’s programs. When he creates his résumé, it helps him a lot to really put himself out there saying, “Okay, these are my strengths, these are the areas where I’ve worked very strongly, and I can be a part of your company and help.” So; that is one area where we use this whole data which we have, in supporting him with the career services we provide to help him find a good job.

Secondly, from our side, machine learning and data analytics also help improve overall learning outcomes. Today, more than 80% of our learners complete their courses; a figure that would be impossible without the personalised, AI-driven approach we’ve implemented. On other platforms, completion rates often struggle to cross 10%. This stark difference underscores the effectiveness of tailoring content, support, and interaction to each learner’s unique journey.

80%+ completion rate is a great metric. Tell us about assessments. Do you use generative AI for automatic assessments or personalised feedback to learners? How does that whole ecosystem work for you?

Yes, definitely, in a lot of areas and we have been playing around with generative AI and large language models; we are employing these tools to assist in evaluating learners in real time. One of the things we have developed is a coding tool known as CodeX.
When a learner is writing a code, it assesses that code in real time. It does not only give a score or output, but also gives them hints on how to improve, so they are able to go back and fix their code. In the long run, this gives them a much better understanding. So we are definitely using GenAI to give that kind of personalised feedback and also to help our mentors. Even mentors are now able to use these tools to give more contextually rich, fast feedback, which otherwise used to take a lot of time.

So, GenAI is helping the learner and also the mentor. It’s creating an ecosystem where assessment is much faster, and feedback is far more effective and meaningful.