Compiler Engineer (AI/GPU)

Location: Paris, France; hybrid work with the ability to relocate to Paris and work closely with the team

Company

KOG builds a co-designed inference stack for real-time AI agents across model architecture, inference engines, compilers, and low-level GPU kernels.

What you will do

• Build AGCO, an agentic compiler for optimizing LLM execution across different GPUs and optimization targets.
• Design compiler and intermediate-representation systems for representing and transforming LLM computations.
• Develop optimization passes, lowering pipelines, and code generation.
• Create search methods that explore alternative implementations and execution strategies.
• Build verification and correctness checks for generated changes.
• Profile and optimize GPU execution, memory, parallelism, communication, and LLM inference performance.

Requirements

• Deep technical expertise and original work in at least one relevant area.
• Experience with compiler engineering, optimization passes, IRs, lowering, code generation, LLVM, or MLIR.
• Experience with GPU programming or performance engineering using CUDA, HIP, Metal, Vulkan, or similar technologies.
• Knowledge of LLM inference engines, attention, MoE, parallelism, communication, or other systems-level LLM execution concepts.
• Experience with formal verification, equivalence checking, SAT/SMT, or systems that automatically generate, search, test, benchmark, or optimize code.
• Ability to explain technical decisions, alternatives, and measured results through code, papers, theses, projects, or detailed technical work.

Culture & Benefits

• Small engineering and research team of 10 people, including 9 engineers and researchers and 4 PhDs.
• Direct collaboration across the full inference stack.
• Fast iteration from optimization ideas to compilation, execution, verification, and measurement on real GPUs.
• High ownership over technical decisions and the systems shaping LLM inference optimization.
• Opportunity to deepen expertise in compilers, GPU systems, or LLM inference while expanding across the stack.

Hiring process

• Technical work is reviewed during the process, including public code, upstream contributions, papers, theses, technical projects, or detailed write-ups.