5+ years of software engineering experience, ideally in customer-facing roles.
2+ years of experience building production-level LLM features.
Strong proficiency in Python, including packaging and testing.
Hands-on experience with MCP servers and LLM tool design.
Familiarity with current agent harnesses and their limitations.
Understanding of agent architecture and decision-making processes.
Experience in running evaluations on LLM systems and interpreting results.
Safety-focused mindset with knowledge of human-in-the-loop systems.
Responsibilities
Own the agent platform from design to support.
Design tools for agents to effectively plan and execute workflows.
Ensure platform portability and comprehensive documentation.
Integrate safety and traceability into the platform from the outset.
Conduct evaluations to inform development decisions.
Implement changes in small, manageable increments.
Collaborate with scientists to translate research into production features.
Benefits
Competitive compensation package.
Stock options plan for shared success.
Access to cutting-edge tools and collaboration with experts.
Hybrid work arrangement with three days in the Boston office.
Opportunities for professional growth and industry engagement.
Impact-driven culture focused on solving complex AI and hardware challenges.
Full Job Description
Position overview
We are looking for an Applied AI Engineer to own the agent platform that lets AI agents run scientific and engineering workflows reliably and safely. Some of these workflows run in lab environments where reliability and safety matter as much as speed.
You will join a small, cross-functional team and be the main owner of the agent platform code. You will scope your own work and make the calls on how it is built. What you will do
Own the agent platform end to end: design, implementation, releases and support.
Design the typed, tested tools that agents use to plan, run and verify work, and decide what belongs in a tool and what belongs in a prompt.
Keep the platform portable and well documented, so it is easy to adopt and extend.
Build safety and traceability into the platform from the start.
Run evals that drive decisions: your own golden task sets, baselines for every release, and comparisons across clients and models, including local ones.
Ship changes in small, reversible steps, with upgrade paths that need no manual work from users.
Turn research experiments into production features, and work with scientists to bring their workflows into the product.
Write ADRs and review code with the rest of the team.
Key requirements
5+ years of software engineering, mostly in companies shipping software to real customers. Research or academic experience is welcome, but it should not be your only background. Software engineers who moved into AI are a good fit.
2+ years building LLM features that run in production at a company, beyond prototypes.
Strong Python: packaging with uv or Poetry, type hints, async, testing and building CLIs.
Hands-on experience building or extending MCP servers and designing tools for LLM tool calling.
Good working knowledge of current agent harnesses (Claude Code, Codex, Cursor or similar): plugins, skills, hooks, subagents, MCP support, and where each one falls short. You use them every day in your own work.
A working understanding of agent architecture: how the model, the harness, tools and state fit together, and which decisions belong in code instead of a prompt.
Experience running evals on LLM systems: building test sets, catching regressions and explaining the results to others.
A safety mindset: human-in-the-loop approvals, guardrails in code, audit logs and awareness of prompt injection.
Care for compatibility: versioning, smooth upgrades and installers you can run twice without breaking anything.
CI/CD with GitHub Actions and Docker.
Comfort in a small team with unclear requirements. You can scope your own work and ship it with little support.
Clear, confident communication with both technical and non-technical stakeholders, including the ability to challenge or push back constructively when needed.
Professional fluency in English.
Nice to have
Experience running LLMs and software in restricted or customer-managed environments.
Pydantic AI, LangGraph or similar agent frameworks.
Lab automation, instrument control or hardware connections (VISA/SCPI, OpenTAP).
Competitive Compensation: A competitive compensation package.
Stock Options Plan: Empowering you to share in our success and growth.
Cutting-Edge Tools: Access to state-of-the-art tools and collaborative opportunities with leading experts in artificial intelligence, physics, hardware and electronic design automation.
Work-Life Balance: A hybrid work arrangement with three days per week in our Boston office.
Professional Growth: Opportunities to attend industry conferences, present research findings, and engage with the global AI research community.
Impact-Driven Culture: Join a passionate team focused on solving some of the most challenging problems at the intersection of AI and hardware.
