Principal Machine Learning Engineer
Edison Scientific7 months ago
Base Salary
$275k - $350k/yr
Responsibilities
- Interpret qualitative scientific AI-agent challenges as well-formulated optimizable problems.
- Build environments for training and deploying AI agents that solve scientific tasks.
- Develop and scale training-data pipelines with observability and reproducibility.
- Lead training of large-scale LLM-based systems and build infrastructure to improve experimentation and production training efficiency.
- Build efficient, flexible inference infrastructure supporting complex sampling algorithms and custom architectures.
- Develop and extend the experimentation platform for internal tools and projects.
- Collaborate with AI researchers, chemists, and biologists on scientific discovery applications.
Requirements
- 8–10+ years of experience in applied ML research and applying ML methods to real-world problems.
- Experience across the ML lifecycle, including data pipelines and provenance, model training, model deployment, and production validation.
- Fluency in PyTorch, JAX, or an equivalent framework.
- Demonstrated experimentation experience in academic or industry settings.
- Strong programming expertise across data, ML, and LLM software stacks.
- A PhD in Machine Learning, Computer Science, or another quantitative field is a bonus.
- Familiarity with leveraging and managing distributed computing resources is a bonus.
- Background architecting complex distributed systems is a bonus.
Benefits
- On-site work at the San Francisco office in the Dogpatch neighborhood.
- Full healthcare coverage with 100% of premiums paid for the employee and dependents.
- Yearly new parent stipend and fertility coverage through Carrot.
- 401(k) company matching.
- $300 health and wellness benefit.
- Free lunch each office day and dinner when working late.
- Regular team offsites and company events.
- Competitive salary and equity.
Tech Stack
Categories
About Edison Scientific
Edison Scientific builds AI scientist agents and an R&D platform (Kosmos) for pharma, biotech, and materials teams, integrating with organizations’ proprietary data to guide experiments and analysis from discovery through development. Headquartered in San Francisco and privately held, it was spun out of the nonprofit research lab FutureHouse in 2025, and offers enterprise deployments—including bring-your-own-cloud installations—to meet compliance requirements at large research organizations.