ML Research Engineer, Performance Optimization

Lila is building a platform where AI and automation co-evolve to accelerate scientific discovery Our AI Research team is looking to hire a ML Research Engineer to maximize iteration speed for mid- and post-training runs on 1T+ parameter LLMs using the latest NVIDIA chips. This involves not only maximizing GPU utilization on an individual training run but also making these optimizations maintainable, reusable, and adaptable so that the rest of the research team always knows how to confidently use the latest optimized stack.

• GPU compute performance: profiling, kernel development, long-context handling, memory footprint optimization
• GPU cluster performance: collective/MoE framework optimization, comm/compute overlap, RDMA/IB/EFA usage, P/D disaggregation
• Big data handling: manage mid-training scale data across storage, streaming, tokenization, and processing
• RL pipeline: optimize the async RL cycle, including environment/sandbox interfacing, long rollout handling, large token metadata (such as router replay), and weight sync
• Setting the performance standard: make sure optimized configurations are tested, accessible, and adaptable to research tasks so that the whole team can work at maximum velocity
• Establishing the Pareto frontier: communicate with researchers about throughput/accuracy tradeoffs (e.g. lower-precision compute, higher async staleness) to establish optimal points

• BS/MS in Computer Science or a related quantitative field and 4+ years working as a software engineer or machine learning engineer, or PhD in a relevant research area and 1+ years work experience.
• Experience with LLM inference libraries such as vLLM and training libraries such as Megatron, including usage and throughput optimization.
• Experience running LLM inference and/or training at scale across multi-node GPU clusters.
• Familiarity with core concepts in reinforcement learning.
• Proven track record of software engineering excellence — clean APIs, well-tested code, thoughtful abstractions.
• Hands-on experience supporting and maintaining production libraries used by other engineers and researchers.
• Strong CI/CD skills — experience designing integration test suites, build pipelines, and automated quality gates for multi-package ecosystems.
• A passion for agentic systems and awareness of cutting-edge approaches in LLM tool use and autonomy.

• Experience working with kernel DSLs and frameworks like CUDA, CuTe, CuTile, Triton, etc.
• Experience with GPU networking frameworks like NCCL and NIXL, as well as fabrics like NVLink, IB, and EFA.
• Experience with distributed systems or compute frameworks (Ray, Flyte, Airflow, or similar).
• Familiarity with containerized compute backends (Docker, Modal) and sandbox orchestration.
• Experience with Ray actors, Ray Serve, or similar actor-model frameworks.

Lila Sciences is committed to equal employment opportunity regardless of race, color, ancestry, religion, sex, national origin, sexual orientation, age, citizenship, marital status, disability, gender identity or Veteran status.

Information you provide during your application process will be handled in accordance with our Candidate Privacy Policy.

Lila Sciences does not accept unsolicited resumes from any source other than candidates. The submission of unsolicited resumes by recruitment or staffing agencies to Lila Sciences or its employees is strictly prohibited unless contacted directly by Lila Science’s internal Talent Acquisition team. Any resume submitted by an agency in the absence of a signed agreement will automatically become the property of Lila Sciences, and Lila Sciences will not owe any referral or other fees with respect thereto.