About Grafton Sciences

We’re building AI systems with general physical ability — the capacity to experiment, engineer, or manufacture anything. We believe achieving this is a key step towards building superintelligence. With deep technical roots and real-world progress at scale (e.g., a $42M NIH project), we’re pushing the frontier of physical AI. Joining us means inventing from first principles, owning real systems end-to-end, and helping build a capability the world has never had before.

About the Role

We’re seeking a Multiagent Systems Engineer to design and implement the agentic reasoning, planning, and orchestration behaviors that enable LLMs to reliably operate complex engineering workflows. You’ll build LangGraph planners, structured state machines, action schemas, and recovery strategies that translate intent into deterministic, traceable tool use. This role is for someone who thrives at the intersection of language models, structured representations, tool interfaces, and real system constraints.

Responsibilities

• Design agent reasoning loops: task decomposition, plan refinement, tool selection, intermediate validation, and stopping criteria avoiding communication frictions.

• Build agent planners, tool-call flows, and state machines using LangGraph or similar frameworks.

• Design schemas, action adapters, and cross-tool interfaces that enable deterministic, traceable, replayable LLM tool use.

• Implement error handling, timeout strategies, retries, rollbacks, and checkpoint/resume mechanisms for robust agent behavior.

• Collaborate closely with systems architecture, data infrastructure, and ML teams to integrate agents into real engineering toolchains.

Qualifications

• Strong experience building agent systems, LLM tool-calling pipelines, or structured orchestration frameworks.

• Experience with multi-agent systems, with a focus on complex communication and orchestration scenarios.

• Research on enhancing reasoning by multiagent systems introduction is preferred.

• Deep intuition for state representations, schemas, action abstractions, reproducibility, and deterministic execution of multi-step workflows.

• Ability to reason about system-level failure modes, edge cases, uncertainty, and safe tool interaction in complex environments.

• Comfortable working across AI, systems engineering, and domain-tooling interfaces in a fast-moving, high-precision setting.

Above all, we look for candidates who can demonstrate world-class excellence.

Compensation

We offer competitive salary, meaningful equity, and benefits.