Practical Tips on Academia-Industry Collaboration: Alt funding

Zhou Yu’s practical tips on how academics can collaborate with industry to explore alternative funding sources.

Author: Zhou (Jo) Yu is an Associate Professor at Columbia University, working on AI.

1. What motivated you to pursue industrial collaborations? What are their benefits?

The motivation comes from the opportunity to align cutting-edge academic research with real-world impact. Especially for people like me, working on AI Agents, LLMs, the industry has great colleges to work with, and has the problem, the data, and computing resources, and the domain knowledge. To me, industry collaboration is a “must have” not a “nice have”.

Industry collaborations offer shorter, more flexible funding cycles than government grants and open doors to valuable applied insights. They are particularly attractive because they’re faster to approve, less bureaucratic, and more likely to evolve into multi-year engagements when both sides find alignment.

2. In your field, what companies have you collaborated with? Why do they fit your research field and your own expertise?

Collaborations have included Amazon, Microsoft, Google, Meta, Intel, IBM, Walmart, Sony,  Tencent, etc—organizations that actively invest in AI, machine learning, and natural language technologies. These companies fit because they either lack in-house expertise in these areas or aim to stay engaged with academic advancements. The work often aligns with research interests in scalable AI systems and open-source innovation.

Recently, Eugene Wu (Database Expert) and I started the Columbia DAPlab, a 5-year lab focused on advancing AI Agent research. We aim to collaborate with a diverse group of industry partners to understand the common problems and build a standardized stack and benchmark to reduce adoption friction.

Together with another 11 faculty members in Columbia, ranging from General ML algorithms faculty, such as Rich Zemel, Daniel Hsu, and Dave Blei, HCI faculty, Lydia Chilton, System faculty, Kostis Kaffes, and Baishaki Ray, Security Faculty, Junfeng Yang, AI Application Faculty, Carl Vondrick, Yunzhu Li,  IEOR faculty, Adam Elmachtoub and Shipra Agrawal, and Business School faculty, Tianyi Peng, DAPlab covers all aspects that support new ideation to system building.

This is modeled after the Berkeley AMP and SKY lab to centralize the industry funding. So individual faculty members do not have to elicit and maintain the industry relationship, but the lab does it together. The 5-year commitment also helps stabilize the funding source for PhDs. We have many industry participants now and are looking for more. Please contact me and Eugene if you would like to be part of it.

3. To establish industrial collaboration, what do you need to do?

It starts with a mix of strategic application (e.g., to open faculty awards) and relationship-building. Submitting to corporate faculty award programs is a great entry point, especially for junior faculty. Additionally, nurturing personal connections, giving company talks, engaging with industry reps at conferences, and maintaining long-term relationships are key strategies.

4. How do you identify and select the right industry partner(s) for collaboration? Can you share a particular example?

Start with people you know—former classmates, intern mentors, or past collaborators who’ve moved to industry. Let them be your champion to help you navigate the organization chart to find the budget holder. Then find the alignment between them and your research team.

I have a long-running collaboration with Sony that began when a previously rejected NSF Career proposal was submitted as a blind application to their faculty award program. That proposal led to a multi-year engagement.

5. During the collaboration, what role do you play, and what role does your industrial collaborator play? What are the expectations of your industrial partners?

Faculty typically lead the research direction and maintain academic quality, including publishing. Industry partners expect regular updates and alignment with their strategic goals. They provide practical feedback, access to datasets or problems, and sometimes co-develop prototypes. A healthy cadence of monthly or quarterly meetings keeps both sides aligned.

6. How do you define success in an industrial collaboration? Can you share a particularly successful industrial collaboration and what made it work so well?

Success is defined by sustained alignment, meaningful joint outcomes (e.g., publications, prototypes), and the continuation of funding over time. The Sony example stands out: a collaboration that started from a cold submission and continues years later, based on mutual respect, technical alignment, and proactive relationship management.

7. What are the biggest challenges you’ve faced in industrial collaborations, and how did you overcome them?

One challenge is ensuring research remains publishable and academically meaningful while meeting industry timelines and goals. Another is navigating the legal bureaucracy of contracts. Always ask for gifts over contracts when possible, and maintain clarity about project scope and deliverables early on.

8. Have you experienced collaborations that didn’t work out? What lessons did you take from them?

Collaborations that don’t align with your research goals or don’t respect academic boundaries may not be worth pursuing. The lesson is to feel comfortable saying no and focusing only on mutually beneficial partnerships.

9. How do you handle IP rights and data sharing in collaborations?

Always prefer gifts which doesn’t involve IP claims or legal encumbrances to awards. It allows faster collaboration and avoids the bureaucratic delays associated with contracts. In AI research, where much of the work is open-source, gifts are especially practical and low-risk for both sides. Many companies have not done this before, so you need to guide them through the process. If you don’t know how to do it, ask a senior faculty member in your department who has received industry gifts before. Every department can be slightly different, but I would get a general gift funding letter template ready and send it to your collaboration partner as soon as possible to kick off their legal review. I cannot offer legal advice, but I can offer tactics to deal with corporate lawyers. 🙂

10. Lastly, what advice would you give to faculty just starting to explore industrial partnerships?

Start early, apply broadly to open calls, and don’t be discouraged by rejections. Put yourself out there. Build relationships with people you already know, be visible at conferences, and share your work on platforms like LinkedIn and X. Most importantly, be intentional—choose collaborations that align with your research and values, and maintain those relationships over the long term. The universe conspires to help the dreamer. You just need to let the universe see you.