Join

What we are working on

Concrete open problems, rather than a description of the field:

• Extending generative binder design to target classes it has not handled well. Small molecules and enzymes are in scope already; the harder cases — shorter peptides, and targets whose chemistry nature never evolved a protein binder for — are open.
• Working out when pretraining actually helps. FLIP2 found that simpler models often match or beat fine-tuned protein language models under realistic splits. That result needs explaining, not explaining away.
• Proteome-scale structure and search infrastructure. Predicting 31 million complexes surfaces engineering problems — long-sequence regimes that fail, recovery of missing predictions, confidence criteria that hold up — that are genuinely research-shaped.
• What complexes show that monomers do not. Emergent folds and topologies that only appear when you predict the pair, and what they imply for function.

Two roles, two doors

Chris is a Senior Research Scientist and Applied Research Team Lead in Digital Biology at NVIDIA, and a Visiting Assistant Professor at Duke University, where he leads this lab.

The Duke lab is independent of NVIDIA. Joining the Duke group does not imply employment at, or access to the resources of, NVIDIA. For NVIDIA roles, see the NVIDIA careers page. Everything below concerns the Duke group.

Projects for Duke students

The lab offers (co-)supervision for students at Duke. To qualify, you must be enrolled in Duke Cell Biology or B&B departments; for co-supervision, we may see case-by-case based on other Duke faculty involved. For the time being, we can offer co-supervision of PhD students and supervision of MSc students.

To request supervision for a project, please send an email to lab at learning.bio with the following:

1. A concise, one-page CV.
2. A research project proposal that aligns with the lab's objectives (250 ± 25 words).
3. A description of your most recent open-source contribution, with links to the code or commits (up to 250 words).
4. Optional: your transcript of records.

The open-source item is not a formality. It is the part we read most carefully.

Projects for non-Duke students, including international MSc and BSc theses

The lab cannot currently sponsor paid positions or visas to the U.S., but there are other mechanisms to collaborate:

• If you have a scholarship, or want to participate in a self-funded project or internship, send an email with the details listed above to the same address.
• The lab offers jointly-supervised internships, theses and projects with international collaborators. We have great collaborators in several universities in the U.K., Germany, Spain and Korea, some of whom are listed on the team page.

Fellowships

Chris and Michael Heinzinger support applications to the Marie Skłodowska-Curie Postdoctoral Fellowship, hosted across Duke, Helmholtz Munich, and the Technical University of Munich. The 2025 call posting describes what the fellowship involves and who it suits; the deadline in it has passed, but the offer to support strong applicants in future calls has not. Write to msca at learning.bio if you are considering it.

What to expect

Work in this lab is open by default — code, weights, and data are released, and you will be expected to write software other people can run. Projects sit close to real collaborators, which means occasional deadlines set by someone else's experiment rather than by us.

You should be comfortable being told that a result is not yet convincing, and comfortable saying the same to us.

On replies

A personal note from Chris, worth reading before you write:

On a regular work day, I receive many more requests through various communication channels directed to me (as in email "to", not CC, not spam) than I could ever possibly tackle in a day, usually asking for some of my time in the immediate future with some level of urgency. My work days could consist entirely in reacting to incoming requests, but this would result in nothing else getting done, which is not a great long-term strategy, and which usually comes at a cost to all ongoing work streams.

I strive to find a balance of replying and addressing requests, and getting work done, which unfortunately means that I won't always reply to all incoming requests, or sometimes be metered in my replies. Please, don't take the lack of a reply personally, or the brevity to heart!

If you are unsure about an answer, consider setting up a call. If on the other hand you have not received a reply after two working days from your request and still need one, send a ping. A sufficient ping is a reply to your original request with "bringing this up to the top of your inbox again".

A few patterns have emerged in what passes the activation (reply) barrier: be direct, be concise, and be personal. If you need something, summarise it in one sentence or a few bullet points at the top of your email — don't be afraid to be direct!

Thank you for understanding,
Chris.

Contact

Write to lab at learning.bio. Please reach out via email for appointments.