18 November 2026
Request for Proposals:
Protein Design and Agentic Biology
The Machine Learning Bio Lab (machine.learning.bio) led by Christian Dallago at Duke develops machine-learning methods to understand and engineer biology. Our work spans protein and genomic sequence models, molecular design, large-scale analysis of protein structures and interactions, and benchmarks that test how well models perform on experimental problems.
We invite applications for a one-year postdoctoral position based in Duke’s Department of Biochemistry and Cell Biology to tackle challenging protein design problems with our collaborators across Duke, and use the results to build better agents and design models. PhD students across Duke who are nearing completion are encouraged to apply.
Location and mentorship
The successful candidate will spend one year embedded in Duke’s department of Biochemistry and Cell Biology, with office space on the 4th floor of Sands at Duke, with day-to-day mentorship from Chris.
Research
You will work directly with experimental collaborators on design campaigns, from defining the problem to interpreting experimental results. Projects may involve selective binders, enzymes or molecular assemblies. The questions and designs will be developed together with collaborators.
The work will use and extend BindClaw and models such as RFD4-Proteina. You will use and adapt agents that run design tools, assess candidates and revise designs using computational and experimental results.
A central part of the research is understanding what makes these systems work or fail. You will collect agent traces, design histories and experimental outcomes, including unsuccessful designs, and use them to improve reusable workflows and build training datasets and benchmarks. Improvements to agents and models should be tested on new targets, with experimental validation and comparisons to existing methods.
Candidate
We are particularly interested in candidates with experience in protein design and agentic systems, strong programming skills, and an interest in building better research software. You should be able to inspect computational results critically and work closely with experimental biologists.
Applicants should have completed a PhD in a relevant field by the start of the appointment.
Application and proposal
Submit a CV and a research proposal of up to two pages, excluding references. Include links to relevant papers, code or projects, your expected PhD completion date and your availability to start.
The proposal should describe:
- Research problem: a difficult protein design problem or a limitation of current agentic design systems encountered in a biological design campaign. Explain why it matters and what existing approaches cannot do.
- Approach and test: how you would address the problem, the role of agents and design models, and the computational or experimental results that would establish success. Identify relevant controls or comparisons.
- Contribution to better systems: what you would learn from design histories, agent traces and experimental outcomes, and how this could improve agents, benchmarks or models.
- One-year scope: the first experiment or implementation, the main technical risk, and a realistic research outcome within the appointment.
Preliminary data and an existing experimental collaborator are not required. The proposal is a starting point; the project will be developed with collaborators at Duke. We will assess the scientific question, technical approach and evidence of your ability to carry out the work.
Send applications to lab <AT> learning.bio with the subject “Postdoc Proposal – Protein Design and Agentic Biology – [Last Name]”.
Application deadline: December 6st, 2026