Research Scientist, Multi-Agent Systems

Plasma AI; New York (NY); On-site; USD 120,000 - 180,000; Full time

Job summary

Plasma is seeking a Research Scientist to study how systems of agents coordinate, communicate, and learn, then integrate findings into Plasma’s orchestration layers. You will publish results for the research community and ship them into Plasma.

The role blends research and product work, requiring hands-on experiments with LLM agents and a strong engineering mindset. A PhD or equivalent experience is a plus.

Qualifications

• Proven research record in multi-agent systems, LLM agents, or related fields such as distributed AI or mechanism design.
• Hands-on experience running experiments with LLMs and agent frameworks.
• Strong engineering: you write the code your experiments run on.
• Rigor about measurement: you check a result before you trust it.

Responsibilities

• Coordinate how agents split a large task and hand off work.
• Design and compare communication methods between agents.
• Test memory, context, and model choices affecting group capabilities.
• Build test environments for multi-agent experiments with shared long tasks.
• Publish results for the research community and operational teams.

Skills

Multi-agent systems

LLM experiments

Agent frameworks

Engineering

Experimental rigor

Education

PhD or equivalent research experience

Tools

Python

Experimentation frameworks

Job description

We're looking for a research scientist to study how systems of agents coordinate. You'll research how agents divide work, communicate, and share what they learn, and build the results into the orchestration layers Plasma runs on.

You've run experiments on LLM agents and want to see the results used by real teams. The role is both research and product: you'll publish what you find and ship it into Plasma.

What You'll Work On

• Coordination: Study how agents split a large task, hand off work, and recover when one of them fails
• Communication: Design and compare the ways agents talk to each other, building on Radio
• Agent Design: Test how choices about memory, context, and model affect what a group of agents can do
• Environments: Build the test environments for multi-agent experiments, where several agents share one long task
• Publication: Write up results for the research community and for the people who run our systems

What We're Looking For

• A research record in multi-agent systems, LLM agents, or a related field such as distributed AI or mechanism design, shown through papers, code, or shipped systems
• Hands-on experience running experiments with LLMs and agent frameworks
• Strong engineering: you write the code your experiments run on
• Rigor about measurement: you check a result before you trust it
• A PhD or equivalent research experience is a plus