Build diagnostic pipelines to identify knowledge gaps
Route failures into data pipelines for continuous improvement
Maintain high-performance infrastructure for LLMs
Benefits
Collaborative, low-ego team environment
Opportunity to work on cutting-edge AI technologies
Focus on continuous learning and improvement
High bandwidth communication through in-person collaboration
Commitment to code quality and rigorous testing
Full Job Description
What You'll Do
Your day-to-day depends on your focus, but the shared surface area includes:
Improve the RSI loop - design and build the core algorithms that drive system-level self-improvement.
Push reasoning strategy - propose and test novel, task-specific reasoning strategies; probe LLMs to surface latent knowledge; build data-efficient methods for finding optimal strategies per task.
Build dynamic harnesses - synthesize task-specific execution environments on the fly, even as the system concurrently rewrites its own codebase.
Architect orchestration - design clean abstractions for highly parallelized workflows, turning fragmented signal from thousands of parallel, asynchronous LLM interactions into coherent, high-performance reasoning chains.
Probe capability boundaries - build diagnostic pipelines that separate genuine knowledge gaps from tooling limits (e.g., live retrieval) and reasoning/composition failures.
Close the feedback loop - route knowledge-gap failures into SFT data pipelines; turn reasoning/composition failures into verifiable, exact-match RL tasks.
Run infrastructure at scale - build and maintain high-performance infrastructure against external, third-party LLMs (experimental, large-scale, and publicly deployed) so research and optimization loops run reliably without human intervention.
Keep it interpretable - every loop stays transparent code, explicit prompts, and clear diagnostics, never opaque parameter updates.
What You'll Need
Everyone on technical staff:
Autonomy to turn both well-defined and highly ambiguous ideas into production-grade reality.
First-principles understanding of LLM reasoning and failure modes, plus hands-on experience with LLM "quirks."
You want to build loops where the system learns from each problem it solves and gets better at the next one.
Fully on-site in Los Altos - this is a high-bandwidth, whiteboard-heavy, low-ego, in-person team, and that's a fit requirement, not a preference.
Commitment to code quality, rigorous testing, performance profiling, deliberate architectural design, and clear communication.
