Job Description, Responsibilities & Requirements
About the Position
We’re looking for a rare type of engineer - someone who can turn cutting-edge AI research into real, production-grade systems. This role sits at the intersection of research and engineering: working directly with model internals while building scalable solutions that are actually used in high-stakes environments.
The team operates at the frontier of open-source AI, developing highly optimized and interpretable models that outperform existing solutions, particularly in regulated industries like finance.
Responsibilities
- Translate mechanistic interpretability research into production-grade systems
- Work directly with model internals to improve performance, reliability, and control
- Apply techniques like activation patching, control vectors, and feature-level interventions
- Build evaluation and deployment pipelines for enterprise-grade environments
- Design systems that enforce deterministic policies at inference time
- Continuously experiment, validate, and ship improvements into production
Requirements
- Strong understanding of Transformer architectures and PyTorch internals
- Solid foundation in deep learning theory and model behavior
- Hands-on experience training, fine-tuning, or optimizing LLMs beyond surface-level approaches
- Ability to read research papers and implement what actually matters
- Experience working directly with model internals (weights, activations, representations)
Nice to Have
- Experience with mechanistic interpretability techniques in practice
- Background working with large-scale or open-source LLMs
- Familiarity with regulated environments (finance, compliance-heavy domains)
- Experience building ML systems that require auditability and strict control
- Contributions to research or open-source projects
- Contributions to open-source projects
Soft Skills
- Strong research + engineering mindset
- High ownership - you don’t ignore problems, you fix them
- Curiosity and ability to quickly absorb new concepts from papers
- Deep intrinsic interest in LLMs and their behavior
- Hands-on approach (you build, not just design)
- Comfort working in a fast-moving, high-trust startup environment
What Makes This Role Different
- You don’t just use models - you intervene in how they think
- You work on systems that must be deterministic, auditable, and reliable
- Your work directly impacts real-world, high-stakes AI deployments
What You Get
- Competitive compensation + meaningful equity
- Direct impact on core AI systems used in production
- High autonomy and trust - strong technical opinions are expected
- Opportunity to push LLMs beyond current limits and define how they behave in the real world
If you are interested, please send your CV!
Work Details
- Work type: Full-time
- Department: Research & Development
- Division: North America
- Location: San Francisco, CA
Apply now
Share to