Postdoc: Formal Verification for ML Inference Engines

École polytechnique fédérale de Lausanne, EPFL; Lausanne; Vor Ort; CHF 90.000 - 120.000; Vollzeit

Benefits dieser Stelle

International working environment

Excellent working conditions

Access to frontier AI models & HPC

Travel funded Lausanne–London

Zusammenfassung

EPFL invites applications for a Postdoc on the Keystone Project (Machine-Verified LLM Inference) in Lausanne. You will contribute to formally verified AI systems, work with Rocq/Lean, and collaborate across an ARIA-funded program with Imperial College London. Strong publication record and independence are expected.

The role offers international collaboration, access to frontier AI models and high-performance compute, and travel between Lausanne and London for joint work and conferences.

Qualifikationen

• PhD (or nearing completion) in computer science or closely related field.
• Background in formal verification, programming languages, systems, or ML.
• Research experience in interactive theorem proving (Rocq, Lean, HOL, Isabelle, or similar).

Aufgaben

• Conduct research for the Keystone project and contribute to a machine-verified LLM inference engine.
• Formalize GPU kernel semantics (e.g., PTX/Triton) in a proof assistant and verify high-performance kernels.
• Implement and verify the inference coordination layer in Rocq/Lean, with extraction to executable code.
• Develop agentic AI workflows for specification autoformalization, proof generation, and proof repair.
• Collaborate with EPFL and Imperial College London teams across visits and open-source releases.

Kenntnisse

Formal verification

Interactive theorem proving

Rocq/Lean

Python

C++

OCaml

GPU programming

English communication

Ausbildung

PhD in computer science

Tools

Rocq

Lean

HOL/Isabelle

Jobbeschreibung

EPFL invites applications for a Postdoc on the Keystone Project (Machine-Verified LLM Inference) in Lausanne. You will contribute to formally verified AI systems, work with Rocq/Lean, and collaborate across an ARIA-funded program with Imperial College London. Strong publication record and independence are expected.

The role offers international collaboration, access to frontier AI models and high-performance compute, and travel between Lausanne and London for joint work and conferences.