Meet ApplyKite Twin, the application assistant that learns with you
Meet your Twin, it learns with you
Your Twin learns with you
Launch week codes:30% off annual Pro20% off monthly Pro
Ends in00h00m00s
Ends in00h00m00s
Nate Foster
Top university
6 days ago
Postdoc: Keystone Project (Machine-Verified LLM Inference) EPFL in Switzerland
Degree Level
Postdoc
Field of study
Computer Science
Funding
Funded postdoctoral position within an ARIA-funded collaboration; 1-year contract, renewable if project duration permits. The advert also mentions funded travel and access to frontier AI models and high-performance compute.
Country
Switzerland
University
EPFL
Apply for this position
Keywords
Suggested positions
About this position
The Keystone project at EPFL focuses on building a formally verified ML inference engine at the intersection of formal methods and high-performance systems. The work combines interactive theorem proving with secure computer systems and ML infrastructure, aiming to make verified systems competitive with unverified ones in development effort, features, and performance.
The postdoctoral researcher will contribute to the design and construction of a machine-verified LLM inference engine. Possible research directions include formalizing GPU kernel semantics such as PTX or Triton in a proof assistant, verifying high-performance inference kernels, implementing and verifying the inference coordination layer in Rocq or Lean, and developing agentic AI workflows for specification autoformalization, proof generation, and proof repair.
The position is part of an ARIA-funded collaboration between EPFL and Imperial College London. The successful candidate will work in a distributed, multi-institution environment with close collaboration with the Imperial team, regular visits to London, engagement with open-source releases of proofs and specifications, and participation in programme activities including red/blue team exercises and sprint reviews.
The role is hosted at EPFL and connects to the LASER group, with access to frontier AI models and high-performance compute for proof-assistant workloads. The advert emphasizes a strong research environment in formal verification, systems, and ML infrastructure, plus collaboration with internationally renowned experts and a team of postdoctoral researchers and PhD students.
Funding details
Funded postdoctoral position within an ARIA-funded collaboration; 1-year contract, renewable if project duration permits. The advert also mentions funded travel and access to frontier AI models and high-performance compute.
What's required
Applicants should have a PhD, or be nearing completion of a PhD, in computer science or a closely related field. The advert seeks candidates with a background in formal verification, programming languages, systems, or ML, and research experience in interactive theorem proving (Rocq, Lean, HOL, Isabelle, or similar), GPU programming or semantics, compilers, concurrency, or ML systems (e.g. vLLM, SGLang, or similar inference engines). Strong computational and analytical skills, solid software engineering ability across multiple languages such as Python, C++, OCaml, and functional languages, a publication record relative to career stage, excellent written and oral English, independence, creativity, solution-oriented working style, and good team spirit are expected. Experience with LLM-based tools for code or proof generation is a plus.
How to apply
Apply through EPFL’s online platform only. Submit a brief cover letter (up to 2 pages), a CV with publication list in one PDF, a research statement (up to 3 pages), and contact details for 3 referees. For questions, contact Nate Foster by email.
Make your next move.
Prepare your application
Review drafts and confirm requirements with the university.
Professors
Professor at EPFL
Switzerland
Research Interests
Computer Science
Programming Language
Information Technology
Formal Methods
Electrical Engineering
Make your next move.
Prepare your application
Review drafts and confirm requirements with the university.