Staff Research Scientist, Behavioral AI & Agents

a16z-speedrun; San Francisco (CA); On-site; USD 130,000 - 180,000; Full time

Benefits offered by this job

Medical, Dental, Vision

Paid time off

Vacation

Job summary

Primitive Labs is seeking a research-focused individual to build the research systems behind behavioral intelligence agents. You will work across simulation architectures, memory systems, cognitive science, continual learning, and agent architectures to improve how agents model human behavior, make decisions, and act within real product experiences.

You will contribute to papers, technical writing, open-source work, and conference talks while helping define our research program.

Qualifications

• Led research in agents, ML, simulation, or cognitive science.
• Design rigorous experiments.
• Build the systems needed to test your ideas.
• Work deeply with large models and agents.
• Turn research into production systems.
• Care about human behavior, not just task completion.

Responsibilities

• Own research directions from question to production.
• Design experiments for agent behavior.
• Build simulations of real user decision-making.
• Develop behavioral-fidelity benchmarks.
• Train and evaluate agents against human data.
• Translate research findings into model, memory, and runtime improvements.
• Publish papers, technical write-ups, and open-source work.

Skills

Research leadership in agents, ML, or.

Experiment design

Systems building

Working with large models

Productionizing research

Human behavior focus

Job description

Primitive Labs is seeking a research-focused individual to build the research systems behind behavioral intelligence agents. You will work across simulation architectures, memory systems, cognitive science, continual learning, and agent architectures to improve how agents model human behavior, make decisions, and act within real product experiences.

You will contribute to papers, technical writing, open-source work, and conference talks while helping define our research program.