Senior AI Compiler Engineer - Applied Research

Overview

NVIDIA's GPUs are at the core of modern AI infrastructure, from training large‑scale models to running inference in production. That position depends on software as much as hardware, and compiler engineering is a big part of what makes it work. The position is full‑time. We are looking for an outstanding AI Research Engineer / Applied Scientist focused on Compilers / Low‑level optimization to join the team and develop groundbreaking technologies in machine‑learning compilers and AI systems.

We build innovative AI compiler solutions that work together with NVIDIA's software stack to provide comprehensive acceleration for modern machine‑learning models.

• Design and implement AI‑based technology addressing core problems of low‑level GPU code generation.
• Build SFT and RL training pipelines.
• Define model inputs using low‑level compiler representations.
• Define, implement, and evaluate strategies for intelligent prompt engineering in the compilation domain.
• Prototype and iterate on model architectures, prompts, and training strategies for NP‑hard problems in optimizing compilers.
• Prepare datasets from compiler traces, optimization passes, and target‑specific performance signals.
• Apply RL techniques to optimize for downstream objectives and run rigorous experiments, analysis, and benchmarking across workloads and hardware targets.
• Build rigorous benchmarks to assess code quality, correctness, and generation overhead.
• Partner with compiler engineers to integrate and ship learned policies with production tool chains.

• M.S. or Ph.D. degree in Computer Engineering, Computer Science or a related technical field (or equivalent experience).
• 5+ years of experience building AI/ML systems.
• Solid understanding of machine learning fundamentals and experimentation best practices.
• Strong software engineering skills in Python and C++.
• Hands‑on experience training, fine‑tuning, and post‑training large models.
• Experience with reinforcement learning.
• Reward modeling from non‑differentiable signals (binary runtime/compile success, performance counters).
• Knowledge of prompt‑engineering techniques (CoT, chaining/orchestration, context adaptation, etc).
• Ability to work across research and engineering, from prototype to production.
• CUDA programming experience and GPU performance familiarity.

• Distributed training/inference at scale (Megatron, NeMo, vLLM, Triton).
• Experience working with the NVIDIA training stacks.
• Fundamentals of construction of optimizing compilers.
• Understanding of GPU performance, experience with benchmarking suites and performance profiling tools.
• Knowledge of formal methods or static analysis for correctness guarantees.

Base salary range: 152,000 USD – 241,500 USD, commensurate with location, experience, and market rates. Employees are also eligible for equity and a generous benefits package.

Applications will be accepted at least until June 20, 2026.

NVIDIA is committed to fostering an inclusive work environment and proudly offers equal opportunities. We do not discriminate on the basis of race, religion, color, national origin, gender, gender expression, sexual orientation, age, marital status, veteran status, disability status or any other characteristic protected by law.