High-Performance Software Engineer — Verified ML Inference Engine

École polytechnique fédérale de Lausanne, EPFL; Lausanne; Vor Ort; CHF 110.000 - 150.000; Vollzeit

Benefits dieser Stelle

Frontier AI models access

Travel for collaboration

Professional development

Excellent working conditions

Zusammenfassung

EPFL seeks a software engineer to help turn verified research prototypes into a production-grade, high-performance LLM inference engine. The role involves designing core components, runtime glue, GPU kernels, and driver interfaces, with collaboration across EPFL and Imperial College London.

You'll organize the engineering infrastructure, run differential testing against reference engines, and contribute to agentic AI pipelines, proofs, and open-source artefacts.

Qualifikationen

• Higher degree in computer science or equivalent; experience in research or industry.
• Strong knowledge of systems programming and performance optimization.
• Experience with one or more of: GPU programming, ML inference serving, formal methods.

Aufgaben

• Design and implement core components of the verified LLM inference engine.
• Collaborate with EPFL and Imperial College London teams.
• Maintain CI and performance benchmarking tooling.
• Develop AI pipelines for specification autoformalization and proof repair.
• Produce documentation and reproducible artefacts.
• Address failures and regressions across the stack.
• Assess security and trusted computing base; contribute to ARIA exercises.
• Contribute to open-source releases and communities.

Kenntnisse

Systems programming

Performance engineering

GPU programming

ML inference serving

Formal verification concepts

Build systems/CI

English communication

Ausbildung

Master's degree in Computer Science

Tools

Python

C/C++

Rust

OCaml

Jobbeschreibung

EPFL seeks a software engineer to help turn verified research prototypes into a production-grade, high-performance LLM inference engine. The role involves designing core components, runtime glue, GPU kernels, and driver interfaces, with collaboration across EPFL and Imperial College London.

You'll organize the engineering infrastructure, run differential testing against reference engines, and contribute to agentic AI pipelines, proofs, and open-source artefacts.