/roles — ROLE_326
Member of Technical Staff, Research
Agent research lab working to make LLM agents reliable, grounded, and useful for people and companies
The role
- COMP
- $280K - $350K
- EQUITY
- Competitive equity (likely in the range of 0.05 - 0.5%)
- LOCATION
- San Francisco · South Bay Area
- WORKPLACE
- On-site
- EXPERIENCE
- 2+ years
- VISA
- None, Visa transfers
- STACK
- Python, PyTorch, TensorFlow
- INDUSTRY
- AI, Software Development
The company
Research lab developing specialized LLM agents that continually learn toward expert-level performance, founded by researchers whose agent work is widely used by frontier labs.
- STAGE
- growth-stage
- FUNDING
- $40M+ raised
- TEAM
- ~10 people
- FOUNDED
- 2025
JD — the work
About the role
You would join the core research team building LLM agents that plan, reason, and take reliable actions in real settings. The lab's goal is agents that are dependable, grounded, and within reach of individuals, developers, and enterprises. You would run research projects yourself, from the first idea and experiments through prototyping and testing products, partnering with designers and engineers so promising ideas turn into practical systems.
What you'll do
- Drive research into post-training, reasoning, and the design of agentic systems
- Develop methods that make autonomous agents more capable, reliable, and safe in real environments
- Partner with engineers to turn research results into prototypes and then into products people keep using
- Run experiments and benchmarks, and study model behavior to uncover failure modes and openings
- Keep up with new work on tool use, RLHF, multi-agent setups, and fine-tuning, and publish or open-source when appropriate
- As an early hire, influence how the lab does research and where its technical roadmap goes
What they're looking for
- 2+ years of research experience in LLMs, agents, or related ML areas
- Strong Python with PyTorch or TensorFlow
- Track record of designing and running experiments end to end
- Familiarity with reasoning, post-training, RLHF, tool use, or multi-agent systems