Postdoc on Machine-Verified LLM Inference (Keystone Project), EPFL
Job Description
EPFL is hiring a postdoc on the Keystone project, machine-verified LLM inference, at the meeting point of formal methods and high-performance systems. The call is open until filled.
At a glance
- Post: Postdoc on Machine-Verified LLM Inference (Keystone Project)
- Where: Keystone project, EPFL, Lausanne, Switzerland
- Pay: on the EPFL scale for this grade; the advert gives no figure
- Closing date: none given. EPFL reviews applications as they arrive, so the post is open until filled. We re-check these listings about every 30 days
Read the official EPFL advertisement before you apply.
The project
Keystone advances two domains together: formal verification and interactive theorem proving with Rocq and Lean, and secure, high-performance computer systems including machine learning infrastructure.
The advertisement seeks candidates in formal methods, computer systems, or both, with interests including machine-checked verification of systems software, and GPU kernel semantics and verification. Verifying the stack that serves a large language model is a genuinely new problem: the kernels are hand-tuned, the numerics are approximate, and the usual proof techniques were not built for either.
What you need
- A PhD in computer science or a closely related field
- Work in formal methods, computer systems, or both
- Interest in one or more of: machine-checked verification of systems software, GPU kernel semantics and verification, and machine learning infrastructure
- Read the official advertisement for the full list of areas
Pay and conditions
EPFL pays doctoral assistants and postdoctoral researchers on published Swiss federal scales, which are among the higher ones in Europe. This advertisement states no figure, so check the scale for your grade and the cost of living in Lausanne before you decide.
How to apply
Apply through the EPFL careers page for this post. The advertisement gives no closing date. EPFL reviews applications as they come in, so apply early.
Browse more research jobs on ResearchJobs.in, or search our Find a supervisor page.
Hiring institution: École Polytechnique Fédérale de Lausanne (EPFL)
Official advertisement: careers.epfl.ch
How to prepare for this application
- You can come from either side. The advert explicitly welcomes formal methods people and systems people.
- GPU kernel semantics is the frontier bit. Say something concrete about it if you can.
- EPFL works in English. Research and teaching at this level run in English, so French is useful but rarely required.
- Doctoral candidates apply twice. A lab offer at EPFL is normally followed by admission to a doctoral programme, so check the programme's own requirements too.
About EPFL
The École Polytechnique Fédérale de Lausanne is one of Switzerland's two federal institutes of technology. It has more than 6,500 staff and over 18,500 people on campus, including more than 14,000 students and 4,000 researchers from over 120 countries, with sites at Lausanne, Neuchâtel, Fribourg, Sion, Geneva and Villigen.
