Founding ML Engineer, RL Environments and Evals

On-site (San Francisco, CA) | $150K to $250K PLEASE INCLUDE YOUR LINKEDIN TO BE CONSIDERED!

About the company

An early-stage AI startup that evaluates hardware engineers on real engineering tasks for companies in space and defense, advanced manufacturing, and robotics, helping them decide who to hire and who to staff on each project.

Spending in these sectors has grown several times over in recent years, and there's still no reliable way to match engineering talent to the work. The product is already live with well-known customers in the space, and the company recently closed its pre-seed round.

The role

You'll be the founding ML hire, taking the company's assessment data (expert-written scoring rubrics and recorded CAD and ECAD sessions from real engineers) and turning it into environments, graders, and datasets that models can train on.

This is a true ML engineering seat focused on reinforcement learning and evals, not an AI applications job built around RAG or prompt tuning.

Your first month leans heavily on data pipelines, making sure what gets built lines up with the data the company produces, including data that frontier labs and fast-growing startups want to use.

After that you'll automate the scoring and question writing the team does by hand today, so the company can take on more demand. Expect the scope to shift quickly as the company grows. What you'll do

• Design and build RL environments and evaluations for agentic CAD and ECAD tasks across mechanical and electrical engineering
• Build graders, reward signals, and verification systems that score engineering work consistently
• Turn assessment data into datasets and environments that frontier AI labs can use in post-training and production training runs
• Own the data pipelines that connect what the company builds to the data it generates
• Automate scoring and question creation so the team spends its time reviewing submissions instead of grading by hand
• Work with the mechanical and electrical engineers who define what correct looks like, and with lab teams who use your environments, then run experiments that shape the research and product roadmap

What you must bring

• 3+ years of full-time ML experience, with core work in RL, evals, data pipelines, or LLM-as-a-judge (rather than mainly computer vision or SFT)
• Real experience with RL environments, either building them for others or using them for post-training
• Grounding in hard tech: ML work at a hardware company, close collaboration with mechanical or electrical engineers, or a mechanical or electrical engineering degree
• ML knowledge that touches hardware, such as CAD, ECAD, 3D vision, or similar
• Hands-on work with evaluations, grading systems, data pipelines, and ML infrastructure
• A track record of taking initiative and owning outcomes without a detailed plan handed to you
• Comfort with ambiguity at a very early-stage company
• Ability to work in person in San Francisco on the schedule below

Nice to have

• ML or AI work at a company building CAD or ECAD software
• Experience with training workflows beyond SFT
• Experience at a fast-scaling startup, or strong post-training work at a large lab
• A history of stepping outside your assigned work and landing on exactly what the team needed next

Compensation & logistics

• $150K to $250K base salary plus 0.2% to 0.7% equity
• On-site in San Francisco, CA, in office 9am to 9pm, six days a week
• Relocation to San Francisco expected if you're not already local
• Not open to visa sponsorship. Candidates must already be authorized to work in the US

This isn't the right fit if

• Your ML experience is mostly applications work like RAG pipelines or prompt engineering
• You have no hands-on exposure to hardware engineering through work, collaboration, or study
• RL is something you've read about rather than something you've built with
• You're looking for an advisory role or prefer work that's fully scoped before you start