Many bachelor's and master's degree students in India aspire to pursue a career in research in AI / ML / Data Science. Unfortunately, those outside the elite institutions rarely have access to a research ecosystem with world-class mentors. In this context, the IKDD Uplink Internship Program is designed with the following goals in mind:

• Provide a paid 3-month summer internship to deserving students, with faculty members at top tier Indian institutes as mentors

• Incentivize excellence in outcome, by providing financial support for publications resulting from this internship work at top tier AI / ML / DS conference and journals

The annual program was launched in 2022. We have been able to offer between 10 and 15 internship positions per edition. Based on our own experience and feedback from students and mentors, the program has evolved over the succeeding editions. The number of applications has grown from around 100 in 2022, to around 300 in 2023, and around 900 in 2024. Among these, the number of high-quality applications has been many times more than the number of available positions. Most potential mentors to whom we reached out agreed with enthusiasm to be part of the program. The majority have returned as mentors in later editions.

Not surprisingly, almost all interns have found the experience to be immensely rewarding. The quality of work, assessed from the final presentations and internship reports, has been high. The major success stories so far are three papers published in CORE A* conferences based on internship work from the first three editions. 

Applications are now open for the 2025 edition. This year we have expanded our focus beyond core AI / ML / DS, NLP and Computer Vision. We have two mentors who specialize in application of AI / ML in biology / genomics. In response, we are looking forward to seeing more interesting profiles in our application pool.

The deadline is 23:59 hours IST Feb 22, 2025.  If you are an eligible intern candidate, do consider applying. (See eligibility criteria here)

• Standouts

• AAMAS 2025: Vibulan J, IIIT Kancheepuram (Intern), Shweta Jain, IIT Ropar (Mentor)

  • "Fair Assignment on Multi-Stage Graphs", Vibulan J, Swapnil Dhamal, Shweta Jain, Ojassvi Kumar, Aman Kumar, Harpreet Singh, Extended Abstract in AAMAS, 2025 

• AAAI 2025: Sadiq Ebrahim, Jadavpur University (Intern), Richa Singh, Mayank Vatsa, IIT Jodhpur (Mentors)

  • “AQUAFace: Age-Invariant Quality Adaptive Face Recognition for Unconstrained Selfie vs ID Verification”, Shivang Agarwal, Sadiq Siraj Ebrahim, Jyoti Chaudhary, Richa Singh, Shyam Prasad Adhikari, Sangeeth Reddy, Mayank Vatsa, Full Paper in AAAI, 2025 

• ACL 2023: Ujan Deb, IIT Bhilai (Intern), Preethi Jyothi, IIT Bombay (Mentor)

  • “Zero-shot Cross-lingual Transfer Using Target Language Projections”, Ujan Deb, Ridayesh Parab, Preethi Jyothi, Short Paper in ACL 2023 (Link)

• Honorable Mention

• NeurIPS Workshop on SafeGenAi 2024: Varun M.S., PES University (Intern), Prathosh A.P., IIsc  (Mentor)

  • "Variational Diffusion Unlearning: a variational inference framework for unlearning in diffusion models", Subhodip Panda, Varun M S, Shreyans Jain, Sarthak Kumar Maharana, Prathosh AP, Paper in NeurIPS Workshop on SafeGenAi, 2024 (Link)

Thanks for your time. In case you are interested in the story behind Uplink, please read on. 

Sometime in the middle of the pandemic in 2021, while brainstorming about new initiatives that the new IKDD Committee could take up, one of the committee members wondered aloud if there was something that IKDD could do to help in mentoring students outside of the small select set of top-ranked institutes in India. This was an idea everyone liked, and that's where the long journey started.

Progress was painfully slow at the beginning. There were questions galore, with few satisfactory answers. Who could be the mentors? How would we incentivize the mentors? What would be the mode of mentorship? Would students be interested in yet another internship program? If we wanted to pay stipends, how many students could we support and how would we get the funds? How would we measure success? We persevered, and slowly but surely clarity emerged. Around January 2022, we had a workable plan though with little certainty of success. We decided to go ahead and execute the plan, while being prepared to abandon it after one edition if we assessed it to be a failure.

Despite all the uncertainty, and the volume of effort involved, why did we decide to go ahead? We had in mind the best-case scenario - some student from a lower ranked academic institute excels in this internship and goes on to publish a paper in a top-tier conference. What would that mean to this student? How might it change the student's career trajectory? What would that mean to this student's peers in the same institute, and in other similar institutes? What myths would that help shatter? This is the vision that inspired us to give it a shot.

We knew that getting mentors to sign up in the first year would be a challenge. No one knew what to expect from the interns. It was a gamble, even less likely to pay off due to the remote nature of the internship enforced by Covid. And we knew that top tier mentors do not typically like to gamble with their time. However, we were pleasantly surprised. While some candidate mentors declined our invitation, most agreed - not for any of the incentives we designed, but because they identified with the cause and wanted to help.

Now that we had mentors on board, we faced the next set of uncertainties. Would we be able to find decent candidates? Was our belief about the untapped talent pool outside of the top colleges well founded? Would we be able to reach out to enough such candidates? As the deadline passed, we looked through the 100-odd applications that we had received, and our concerns were put to rest. We had asked for a short statement of purpose to be submitted along with the application, explaining why the candidate felt this internship was necessary, what made the candidate suitable and what the candidate hoped to get out of this opportunity. The high quality of some of the CVs aside, so many of the SoPs were eye-openers. These were lessons on striving in the face of adversity to make something meaningful out of life. That we would not be able to accommodate most of these candidates given our limited slots was saddening.

As the internship kicked off, we regularly polled both the candidates and the mentors for feedback and were relieved to learn that all was going well. While there were a couple of mentors who felt that the interns fell below their expectations, most were satisfied, and a few were ecstatic about their interns.  Before the year turned, one of the interns had produced an ACL paper.

After the success of the first edition, the going expectedly got smoother. For the second edition in 2023, mentors were more willing to sign up, and the number of applications grew three-fold.  We now had a happy problem. How would we go through so many applications? We did not want to trust any automated filter, such as a CGPA-based cut-off. Going through all applications manually was hard work, more so due to the short time window. But when the mentors were satisfied with the personalized shortlists that we prepared for them, that was sufficient reward for us.

The happy problem intensified another three-fold in 2024 - we received close to a thousand applications. Manually assessing all of these seemed nearly impossible, but we decided to dig in and managed to pull it off, again. We found that the volume of high-quality applications had grown proportionately, which unfortunately meant that we were able to accommodate an even smaller percentage of these than before. On the flip side, the overall quality of the selected interns had also gone up, as evident from two top tier papers that resulted from this edition.

For 2025, we have expanded the portfolio of projects from core AI / ML / DS and its applications in Natural Language Processing and Computer Vision, to include applications of AI / ML / DS techniques in the natural sciences. We are eagerly looking forward to more surprises, of the pleasant sort, in this edition.

There are obvious ways for the program to improve. 

• The current scale of the program is miniscule compared to the size of the talent pool. However, there are several challenges to scaling up. 

• We are constrained by our budget. Having one or more sponsors would enable us to open up more positions. We are working on this.

• Our current success metric is top tier publications in AI forums. However, there are other dimensions to success, publications in top forums of other disciplines (such as the natural sciences), and creation of useful open-source software for important tasks being two examples.

• We would love to have a bigger pool of mentors. But that is easier said than done. While there are many capable mentors in Industry in India and in academic institutes outside India (of Indian origin), enlisting their services has potential undesirable side-effects. We will keep thinking more about this.