Director / Senior Director, Research Engineering, Life Sciences AI

Lila Sciences Lila Sciences · AI Frontier · San Francisco, CA · AI

Director/Senior Director of Research Engineering for Life Sciences AI. This role involves architecting and building the core platform and codebase that scientist-owned models plug into, setting engineering standards, and anticipating infrastructure bottlenecks. It combines hands-on IC work with growing management responsibility, including building and leading a team. The role requires full ML lifecycle expertise and experience taking research code to production, with a focus on enabling rapid iteration while maintaining quality and reproducibility.

What you'd actually do

  1. Design, build, and own the foundation of the LSAI codebase, the core infrastructure that scientist-owned models plug into.
  2. Set and enforce engineering standards for code quality, testing, versioning, documentation, and repository structure.
  3. Make architectural decisions that balance rigor with the reality that most contributors are scientists first, engineers second.
  4. Anticipate infrastructure bottlenecks and define the platform roadmap to enable rapid iteration while maintaining quality and reproducibility.
  5. Build and manage an engineering team over time while remaining a primary hands-on contributor.

Skills

Required

  • Designing and building core software platforms or frameworks
  • Deep platform architecture expertise
  • Building infrastructure and tooling alongside scientists in a research environment
  • Full ML lifecycle expertise across data, training, evaluation, and MLOps
  • Track record of taking research code to production
  • Hands-on building and strategic leadership
  • Mentoring people and setting technical practices

Nice to have

  • Biological applications such as protein design, nucleic-acid design, or cell foundation models
  • Experience in a bioML lab or scientific computing environment
  • Familiarity with or curiosity about computational biology, protein modeling, or ML-adjacent codebases
  • Comfort working around model builders, even if you do not build the models yourself
  • Performance engineering experience, including profiling and optimizing training and inference
  • Experience writing or tuning CUDA or Triton kernels
  • Ability to reason about GPU utilization, MFU/HFU, memory bandwidth, and kernel-level bottlenecks
  • Deep expertise in the modern ML systems stack, including PyTorch internals, mixed precision, and distributed training across multi-GPU or multi-node clusters

What the JD emphasized

  • core codebase
  • core platform
  • scientist-owned models plug into
  • research code to production

Other signals

  • Owns and builds the core codebase that models plug into
  • Architect and build the core platform
  • Full ML lifecycle expertise across data, training, evaluation, and MLOps
  • Take research code to production