(Special Note: If you do not meet the official requirements for this position, i.e. a PhD and 8 years of combined educational+professional experience, but still feel that you would be an excellent fit for this role, please send me an e-mail).
Why should I do a postdoc/research scientist position?
A postdoc/research scientist position is typically a role that folks take on after graduate school, on their way to building their portfolio to support a successful academic faculty or industrial research application. In North America, faculty in the sciences tend to spend an average of 4 years in postdoctoral positions becoming faculty, and I myself spent 5 years. I have the resources and experience to help motivated applicants develop their research portfolios, and turn their research into transferable credits (like top-tier papers, or usable software tools) to help build their careers.
What is this position?
This is a research position in a DARPA-funded AI-for-science project. One of the main goals is to explore how we can determine if an experiment or scientific result is likely to be feasible/work out, before we run that experiment, or to determine if a scientific claim someone else made is likely to be true. This has lots of potential real world applications (like saving time/cost/resources, and speeding the pace of scientific discovery). There is considerable latitude in how we explore this, but one of the approaches we explore should be using automated (code-based) experiments that give evidence that support something likely being true/feasible (or not).
What would help the person be successful?
Here are the items I look for:
- PhD in computer science, or a related field.
- Past background in AI, and ideally NLP too
- Ideally the person has worked with agents, code generation, extracting things from scientific literature, or some combination of those in the past — or has some convincing story of how their set of skills would be great for this task.
- Experience publishing papers in top AI or NLP conferences is prefered.
- AI generated code is everywhere. The person needs to be able to show code, that they themselves wrote (without an LLM), to demonstrate their fluency in computer science.
Potentially Relevant Papers
- Matter-of-Fact: A Benchmark for Verifying the Feasibility of Literature-Supported Claims in Materials Science (EMNLP 2025; https://aclanthology.org/2025.emnlp-main.203/ )
- CodeScientist: End-to-End Semi-Automated Scientific Discovery with Code-based Experimentation (ACL Findings 2025; https://aclanthology.org/2025.findings-acl.692/ )
- Generating Literature-Driven Scientific Theories at Scale (ACL 2026; https://arxiv.org/abs/2601.16282 )
- CodeDistiller: Automatically Generating Code Libraries for Scientific Coding Agents
(ACL 2026; https://openreview.net/forum?id=M7TNf5J26u) - DiscoveryWorld: A Virtual Environment for Developing and Evaluating Automated Scientific Discovery Agents (NeurIPS Spotlight 2024; https://allenai.github.io/discoveryworld/)
- The Scientific Contribution Graph: Automated Literature-based Technological Roadmapping at Scale (submitted; https://arxiv.org/pdf/2605.15011 )
Official Application Link
The position details are below.
Application: https://arizona.csod.com/ux/ats/careersite/4/home/requisition/25414?c=arizona
Remote work: Possible
Start date: We’re hoping to find someone as soon as possible.
FAQ:
Q: The position lists “8 years of experience”
A: My understanding is that this includes educational and industry experience, and that most folks with a PhD would have at least 8 years of experience from their education. If you think you’re a good fit but worried about this issue, send me a note. There is potentially flexibility for the right candidate.