How to build a sustainable research ecosystem

India lacks a good research ecosystem. In the last 30 years, an amazing tech and VC ecosystem has been developed in the country but we cannot simply rely on knowledge transfer alone to boost our economy. In the late 2000s, the smartphone revolution occurred in US and when SOPs to create new apps were developed, India became a global hub of app developers, then the cloud and ecommerce revolution happened and India again became a hub of scaling up these technologies. This pattern repeats often where a technology is developed in America and then scaled to great heights in India. We are currently in the AI revolution and on similar lines, some amazing AI native apps are being developed out of India but not the base structure of the research and development. Although this has worked in the past, most of the monetary value in the tech sector lies at the core of new technology development. Not just in tech but all other fields too, we need to start looking at developing a prudent way to transform into a knowledge creation economy and that starts with developing a world class research ecosystem.

“Scientific research is the main engine of economic growth” (Yann LeCun, 2025 blog) because it develops a constant stream of new technologies, ideas and talent that is needed to solve the most important problems in the world. This in turn attracts better talent, fosters deep tech startups and economic growth.

Developing this kind of a research ecosystem is easier said than done, because research talent operates like a true market without any artificial constraints like work visas or passports. A PhD student or faculty are the most sought after talent and they live in a passport-less world and hence need to often be handpicked and recruited from across the world like China’s “1000 talents program”. On top of this, you also need a constant source of funding which has a really high risk capacity

On a public policy side, it is a difficult challenge to solve but the lack of research labs and scientists also shows a gap in the Indian deep tech ecosystem which is potentially a great opportunity for a company. There is an entire pillar of research talent, faculty and labs that are missing in India which is a great point to build from.

This is what the iceberg of good science and tech looks like. You require multiple levels of support from both public and private entities to solve meaningful scientific problems. In order to build a healthy research ecosystem in India, we cannot simply replicate this model but must also understand the pain points within it so we can build a better ecosystem that will attract global talent. I want to use my experience to further dig deep into the problems of the current research ecosystem.

As an ML scientist having worked in multiple startups in research labs, there are a few problems I see in this space:

• Lack of funding: PI’s must spend too much time thinking about funding, applying for grants

• Researchers often chase publications. This is too short a dream, you need to be chasing a novel idea or solve an impactful problem that will open up the field more. A paper should only be a byproduct of your work but instead its often what people start chasing leading to incremental work with beautifully designed figures. That’s why most of research papers are BS. This still happens very often because a research paper in credible journal is the only way a researcher or a PI can prove their credibility to their university. Working on hard problems requires longer term commitment and that might mean you may not get some important publications but might still be pushing science.

• Researchers are great at solving problems but suck at finding the right ones to solve. Research problem discovery is given very little time as compared to actually solving the problem. What problem to work on is one of the most important questions to be answered but researchers often pursue projects that is simply cool without the longer term thinking. Figuring out a methodical way of finding an important research project is still an open problem.

• Lack of diversity: 20 CS majors are going to roughly gravitate towards a unidimensional approach so you need a more diverse group of people with varying backgrounds and experiences to be able to attack the problem from multiple angles. A lot of research happens in interdisciplinary fields which requires a team with diverse skill sets that many labs do not have.

Almost all of these problems can be solved by having a strong industry — research lab partnership. Any significant partnership between a lab and a company benefits both entities. Companies can fund longer term research goals that would help their future products. On the other hand, it helps the lab gather funding for its research and find direction of problems to be solving. If a research lab can make this source of funding its revenue source than it can become possible to create a sustainable way to build a research lab. There are some kinks in this system but we can figure this out by learning from others.

Stanford does this through its Human Centered AI institute. Many large conglomerates across the world pay to be an industry affiliate at the HAI institute. They pay $1M / year for the partnership and in return they:

• Work with the team to figure out the best faculty to collaborate with.

• Determine a research project scope and invite faculty to provide project proposals.

• Work with dedicated PhD students and faculty throughout the year to meet expectations.

• Company employees to come work as visiting scholars

• $550k worth of research credits which can be spent directly on any project they like.

• At the end of the year, all research is published, adding to open scientific development.

• These lasting partnerships result in new projects and give an access pool of talent and faculty to the company to collaborate with.

I reckon there is a way to replicate this on a smaller level by starting an AI lab that provides such partnerships to tech companies in India and US. Choosing the right company and ensuring that your end goal is research so you do not turn into a consulting firm is important.

To understand this better I want to speak with research lab heads to understand how they run their labs. I will be speaking with:

• Eugene Wu, a Prof at Columbia who is starting his new lab this year with a strong focus towards partnerships with industry.

  Questions for Eugene Wu

• Jia-Bin Huang, UMCP AI professor who started content creation and also teaches a course on how to conduct research

• Paras Chopra, founder of LossFunk a new AI research lab in Bangalore

  https://letters.lossfunk.com/p/what-is-research-and-how-to-do-it

• David Levi and Marc Gough, manager of industry partnerships at Stanford HAI

• Saad Nadeem, MSK Lab

• Iain Carmichael, ofc! Amazing digital pathology professor and a great mentor.

Important questions for AI research labs!

I also need to talk with a lot of industry research lab heads to get an actual perspective from different folks

• NVIDIA 3D Ai lab, edify team

• Naushad Forbes, Co-Founder Forbes Marshall, was a panellist at Columbia’s India event, has started a research led university called Nayanta, LinkedIn - Innovation in Indian firms

• Manish Gupta - Principal AS, Microsoft Bangalore

Lastly, I plan to use Chad Vanderbilt’s lab as a testing ground for developing new partnerships with startups that can help us direct our research and also give some money to the lab.

Future concerns:

• In economies, some functions and operations are left to the markets to solve, for some others you need the govt. Things like climate change or ending poverty does not have enough incentives in the markets to be working on and needs the collective support of govts to be solved. Similarly, there is some research that needs to be worked on by companies for which they have the right incentives and other kinds of more fundamental research needs to be left to the universities / govt grants to be supported from. To get more technical, research is often broken off into technology readiness levels, and TRL 1-3 (basic fundamental sciences) will require more govt support whereas TRL 7-9 (prototype demonstrations) can be left to big tech / startups.

• Conversations with all these research lab heads is really valuable and needs to be documented well for anyone looking to start a research lab