Acceler8 Talent
Senior LLM Research scientist
Acceler8 Talent, San Francisco, California, United States, 94199
What if your next role lets you push the frontier of LLM planning and tool‑use—building agents that don’t just talk about the physical world, but act in it to engineer, experiment, and manufacture autonomously?
Our client—a hyper‑talent‑dense physical‑AI startup backed by top‑tier builders across frontier AI and deep tech—is hiring a
Senior LLM Research Scientist (San Francisco / Redwood City, In‑Person) . They’re building
Physical Superintelligence : AI systems with general physical ability—the capacity to experiment, engineer, or manufacture anything. Instead of only learning from static data, their models generate new data through automated real‑world experimentation at massive scale. The role:
Develop planning, reasoning, and structured tool‑use models
that power autonomous engineering agents Lead SFT + preference optimization work
(RLHF / DPO) to push capability and reliability ️
Build verifier‑guided RL + safety constraints
for robust long‑horizon agent behavior Design tool‑calling schemas, recovery logic, and policy constraints
so agents can operate real toolchains Partner across agent systems, data infra, simulation, and tooling teams
to train and evaluate in production workflows ⚙️
High‑ownership research + systems integration
on core models and agent behavior Real‑world impact at national scale —accelerating physical discovery, not just model training A uniquely scaled platform for learning
with automated experimental facilities generating new feedback loops Competitive salary + meaningful equity Tiny, elite team with extremely high technical bar and low ego In‑person, high‑velocity environment
split across SF + Redwood City Seniority Level
Mid‑Senior level Employment Type
Full‑time Job Function
Science Industries
Research Services
#J-18808-Ljbffr
Senior LLM Research Scientist (San Francisco / Redwood City, In‑Person) . They’re building
Physical Superintelligence : AI systems with general physical ability—the capacity to experiment, engineer, or manufacture anything. Instead of only learning from static data, their models generate new data through automated real‑world experimentation at massive scale. The role:
Develop planning, reasoning, and structured tool‑use models
that power autonomous engineering agents Lead SFT + preference optimization work
(RLHF / DPO) to push capability and reliability ️
Build verifier‑guided RL + safety constraints
for robust long‑horizon agent behavior Design tool‑calling schemas, recovery logic, and policy constraints
so agents can operate real toolchains Partner across agent systems, data infra, simulation, and tooling teams
to train and evaluate in production workflows ⚙️
High‑ownership research + systems integration
on core models and agent behavior Real‑world impact at national scale —accelerating physical discovery, not just model training A uniquely scaled platform for learning
with automated experimental facilities generating new feedback loops Competitive salary + meaningful equity Tiny, elite team with extremely high technical bar and low ego In‑person, high‑velocity environment
split across SF + Redwood City Seniority Level
Mid‑Senior level Employment Type
Full‑time Job Function
Science Industries
Research Services
#J-18808-Ljbffr