From teaching researchers how to assemble GPU machines to unpacking the latest multimodal AI models, IIIT Hyderabad’s Summer School on AI has mirrored and helped shape the evolution of AI itself. Here’s a look back through the last decade.
When artificial intelligence exploded into the mainstream, most universities found themselves scrambling to keep pace. Course curricula lagged behind the breathtaking speed at which research was advancing. Faculty were learning alongside students, industry was trying to figure out how to adapt, and young researchers were hungry for guidance.
IIIT Hyderabad saw the challenge before most others did. In 2016, the institute launched what would eventually become the Summer School on AI – a two week-long immersive programme designed not merely to teach emerging technologies, but to help the country’s AI community navigate an entirely new research landscape.
“The AI wave was coming, and we found that universities were not ready,” recalls Prof. C.V. Jawahar, Centre for Visual Information Technology (CVIT), who conceptualised the programme. “We thought we should train people and make them comfortable with this new world.” That simple objective has since grown into one of India’s longest-running and most respected advanced AI schools, with thousands of applicants every year, alumni spread across academia and industry, and speakers drawn from some of the world’s leading research labs.
The Evolution
The first editions of the summer school looked very different from today’s programme. Back then, deep learning itself was still finding its footing. GPUs were unfamiliar territory for many researchers. Frameworks were fragmented. There was no universally accepted toolkit, no established best practices, and certainly no ChatGPT. Rather than offering conventional classroom lectures, the organisers focused on helping participants learn by doing.
“In the first summer school, we even had sessions on assembling a GPU computer,” says Prof. Jawahar. “People weren’t using GPUs then. We had online laboratories, practical sessions, discussions on programming frameworks and research practices. When everything is changing rapidly, people want to know what papers to read, what methods to trust and what the right way to do research is.”
The programme initially catered almost exclusively to PhD and post-doctoral scholars. In fact, the organisers deliberately kept undergraduate registration fees higher, not to generate revenue, but to discourage beginners. “We wanted serious researchers,” chuckles Prof. Jawahar. “Initially, PhD students paid less while undergraduate students paid more because we knew everyone would want to join.”
The strategy worked. Over the first few years, the school became a meeting point for doctoral students from IITs, IISc and leading universities across the country. Many of those participants today occupy faculty positions in IITs and other premier institutions.
Growing with India’s AI ecosystem
As AI education matured across India’s top institutions, the summer school evolved too. Today, participants increasingly come from second- and third-tier universities, emerging engineering colleges, part-time research programmes and smaller institutions eager to build AI expertise. Alongside them are faculty members updating themselves on cutting-edge research, startup founders, industry practitioners, and even experienced professionals looking to transition into AI research.
“When I look at the audience today, I see undergraduate students, PhD scholars, professors and even people in their forties and fifties,” says Prof. Jawahar. “That mix makes the discussions very interesting.” From just 33 external participants during its inaugural edition, the programme has grown steadily to host approximately 200 external participants in its 10th edition. In addition, the summer school regularly includes around 75 nominated students from IIIT-H and nearly 25 invited participants, making it a vibrant platform for academic exchange and collaboration.
Despite its name, the summer school has never been an introductory course. Applicants are therefore carefully selected from thousands of applicants every year, based on their research interests, prior work, publications and academic recommendations. “We don’t want someone who simply says, ‘I want to learn AI from scratch.’ This is for people who are already working in computer vision, machine learning, NLP or related areas.” That said, the professor adds, “I tell participants they may understand only fifty percent of what they hear and that’s perfectly okay. This is about exposing people to cutting-edge research. You become familiar with the broad ideas, and then you go back and read more.”
From Deep Learning To Multi-Modal
Just as AI itself has transformed over the past decade, so has the curriculum. The earliest editions focused on the foundations of deep learning, machine learning and computer vision. As natural language processing gained momentum, it found its place in the programme. Today, the agenda reflects the frontier of AI research.
Large Language Models, Vision Language Models, multimodal AI, foundation models and emerging research challenges dominate the conversations. “We always update the programme according to where the field is heading,” says Prof. Jawahar. “Earlier, we spent much more time on the basics. Today, perhaps only the first day covers foundations. The remaining days are about cutting edge research.”
A Who’s Who of AI Research
The summer school’s reputation owes as much to its participants as to its speakers. Over the years, some of India’s leading AI researchers have shaped the programme alongside Prof. Jawahar. Long-time collaborators such as Prof. Vineeth N. Balasubramanian, IIT Hyderabad, Prof. Vinay P. Namboodiri, University of Bath, and Prof. Chetan Arora, IIT Delhi, have helped identify themes, recommend speakers and build an academic network that spans institutions and industry.
As someone who has had the privilege of being associated with the Summer School since its inception, both as a co-organizer and as a speaker, Prof. Balasubramanian finds it deeply rewarding to be part of this sustained effort “to bring together strong foundations and the latest advances in AI research for a carefully selected audience drawn from academia, industry, and startups, while also featuring outstanding speakers from across the field.”
Every edition brings together faculty from premier institutes, researchers from global companies, startup founders and international experts. “I genuinely look forward to the Summer School every year because speaking here is unlike speaking at most other events. The participants are highly curious, intellectually engaged, and eager to discuss research in depth. That combination creates an atmosphere of excitement and discovery that is energizing for the speakers as much as for the audience. Perhaps the Summer School’s greatest strength is that it inspires participants not just to learn today’s AI, but to contribute to tomorrow’s, and this leads to stimulating discussions,” notes Prof. Subramanian.
“Many speakers come because they know the quality of the audience,” agrees Prof. Jawahar. “At the conclusion of his talk in the latest edition, one of the speakers who is a professor told me he wished he had stayed for two more days after seeing the participants. That kind of engagement matters.”
In Able Hands Of Student Army
The student volunteers are the backbone of the CVIT Summer School, playing a crucial role in ensuring the smooth execution of every session. “Beyond logistics and technical support – where they create and manage a dedicated Discord community, the volunteers also conduct the hands-on laboratory sessions following each expert lecture. These sessions are carefully designed around the topics covered by the speaker on that particular day, allowing participants to immediately apply the concepts they have learned in a practical setting,” explains Ram Sharma who has been on the organizing team since 2017. “For attendees, the hands-on sessions help ground the topics discussed during the day and deepen their understanding. For students, it is a miniature academic experience – an opportunity to design and deliver an academic curriculum. Such experiences hold the potential to kindle a yearning to consider academia as a career,” observes Prof. Ravi Kiran S, CVIT.
Food For Thought
Although the hands-on component has naturally reduced over the years, thanks to accessible tools like Jupyter notebooks, Google Colab and cloud computing, the Summer School remains an intensely immersive experience. Days begin with breakfast and continue until dinner, packing lectures, discussions, networking and informal interactions into a schedule that stretches from morning till night. “I’m a big fan of providing breakfast,” Prof. Jawahar smiles. “If everyone eats together, we can start on time.”
The week has always been about much more than classroom learning. The earlier editions were famous for their elaborate hospitality. Ice creams, generous meals and endless snacks became something of a tradition. “People used to complain that they were gaining weight after spending a week here,” laughs Prof. Jawahar, adding, “We’ve reduced the food a little now.”
The Pandemic Period
“Previously we had one week of Summer School on Computer Vision and another week as Summer School on Machine Learning. In fact there were even preparatory “pre-schools” to help participants bridge knowledge gaps. Now we have combined it into a single week-long Summer School on AI,” says Program Coordinator Aradhana Vinod. “Everything is combined because that’s where the field has moved.”
The pandemic marked one of the programme’s biggest turning points. After a brief pause in 2020, the Summer School reinvented itself as an ambitious month-long online series in 2021 and 2022, bringing together leading researchers from around the world. Evening lectures, weekend lab sessions conducted over Zoom breakout rooms, and internationally renowned speakers made the virtual format unexpectedly successful, attracting 500 to 600 participants. Far exceeding the numbers attracted through the in-person programme typically hosted. “We were honestly quite unsure whether an online Summer School would work. The interaction and networking were a big part of the in-person editions, so there was a lot of initial scepticism,” remarks Cecilia Antony, who was involved only with the online edition. But going online actually worked in the Centre’s favour. “We could invite around 25 speakers from across the world (instead of the usual 10 -12), including Prof. Andrew Zisserman from Oxford!”
Even after returning to campus, some of those virtual lessons endured. Today, the Summer School has become a compact, intensive week, compressing what was once spread over fifteen days without sacrificing content. “Everything gets covered in those seven days which previously used to get covered in fifteen days,” says Aradhana. “It’s jam-packed.”
The audience has evolved just as dramatically. What once primarily catered to senior undergraduate and postgraduate students now attracts first- and second-year students who arrive with far stronger AI foundations than their counterparts a decade ago. “If you compare the early batches and today’s first-year students, they are already well-versed. They learn on their own through online courses,” says Aradhana.
It Takes A Village
Some things have remained remarkably constant. The programme continues to draw over 200 in-person participants every year despite the explosion of freely available online AI courses. For Silar Shaikh who has been a core member of the organizing committee since the school’s inaugural edition, that consistency speaks volumes. “Even after so many freely available online courses, we have never seen a decrease in participation. It’s always 200-plus. That is a sign of our success,” he says.
Much of that success is powered by a large, often unseen team. Faculty organisers, administrative staff, finance personnel, technical teams and even IIIT-H’s own students work in concert for months before the event begins, ensuring that hundreds of participants have a seamless experience. “It would be impossible for faculty members to manage all of this alone,” says Prof. Jawahar. “The team has been phenomenal.” Sujata Vaggarani, a remote staffer’s description of the event says it all. “It is like a festival for us,” she remarks listing out the emotions that the AI gala evokes – not just of tradition and culture but community and togetherness too.
Advancing AI Education and Outreach
Perhaps the programme’s greatest success lies in what happens after participants leave campus. Prof. Balasubramanian confesses that the Summer School has benefited him personally. “I have met and later recruited several of my current and former students through this program. It has become a valuable venue not only for learning and exchanging ideas, but also for discovering talented young researchers,” he says. Prof. Jawahar describes meeting researchers at conferences around the world who introduce themselves with a familiar line: “I attended the Summer School in 2018.” Many have gone on to pursue doctoral studies, publish in leading conferences or become faculty members themselves. “I came here to decide whether I should pursue research. After this Summer School, my interest has only grown stronger,” wrote a participant at the latest edition. Another said, “I attended last year and this year both – the quality keeps improving, especially the topics and lectures.”
This growing alumni network is a quiet reminder of the school’s original mission – not merely to teach AI, but to build a community capable of advancing it. It is only fitting then that in recognition of Prof. Jawahar’s passion for education and training at scale he was awarded the ACM India Outstanding Contribution to Computing Education (OCCE) 2021 Award.
Sarita Chebbi is a compulsive early riser. Devourer of all news. Kettlebell enthusiast. Nit-picker of the written word especially when it’s not her own.