- Set overall code culture and tooling for a fast-growing org
- Help to solve our core technical challenges across verticals. Examples include:
- Docent:
- High-concurrency container-based evals with quick ability to iterate on interventions to agentic trajectories
- Deterministic sandbox execution of code that can efficiently restore state from checkpoints
- Interpretability:
- Inference stacks that are as performant as vLLM but flexible enough to allow complex model introspection and intervention, steering, configurable sampling, etc., and that can scale to 400B+ parameter models
- Behavior elicitation:
- Distributed RL training and roll-outs allowing thousands of concurrent rollouts across machines
- Build great internal tools to speed up the team
- Docent:
- Help tone-set in the organization around best practices for building and path-set on what infra we should build
- Help other team members think through infra challenges
- Exceptional programmer fluent in Python
- Bare metal optimization: know GPUs, other accelerators in and out (low-level performance + optimization + parallel programming)
- Experience engineering at scale (distributed systems, reliability, architecture design)
- Leader on global code quality and health (designing good primitives, managing complexity and scale)
- Bonus: can set up LLM pipelines, e.g. multiple specialized LLMs interacting with each other in a performant and reliable way
- Bonus: experience with open-source community management
What We Do
Transluce is an independent research lab that builds open, scalable technology for understanding AI systems and steering them in the public interest. Transluce means to shine light through something to reveal its structure. Today’s complex AI systems are difficult to understand—not even experts can reliably predict their behavior once deployed. Given AI's extraordinary consequences on society, we need scalable and open analyses of the capabilities and risks of AI systems. We are building open source, AI-driven tools to understand and analyze AI systems. We will apply these tools to open-weight models, so the world can vet our analyses and improve their reliability. Once our technology has been vetted, we will work with frontier AI labs and governments to ensure that internal assessments reach the same standards as our publicly vetted procedures. Email: [email protected]
