Applied ML Engineer, Physical Systems - Boston - $350k + equity

Doc Brown: "Roads? Where we're going, we don't need roads."

Most machine learning jobs end at the dashboard. This one starts where the model meets a building full of running hardware, and someone has to explain why the readings and the predictions disagree.

An early-stage, well-backed AI company with a serious scientific pedigree is hiring an engineer to work directly with the operators of large-scale industrial and compute infrastructure. You will spend real time on their sites, learn how their systems actually behave, and build the ML that earns its place in the operating loop. Less publishing, more deploying.

The Work

• Sit with customers on site and apply ML to genuinely hard physical problems. Predictive models, simulation, digital twins, and automated optimisation loops, delivered as software that runs in production
• Build the prototypes and bespoke tooling that make the capability real for a customer. Working systems and live dashboards, not slide decks
• Turn messy operational reality into usable model inputs: sensor streams, design documentation, maintenance and operating logs, external data feeds. Design the schemas, build the ingestion, get to first signal in days rather than months
• Run engagements end to end. Scoping, build, customer communication, and the judgement call on what to keep. Each deployment should either leave behind something reusable or be wound down honestly

You Might Be Our Person If...

• You have several years building production ML systems, with real grounding in applied physics, computational science or engineering
• You are properly literate in at least one quantitative physical domain. You can read a technical paper in it, hold your own with an experienced domain engineer, and reason honestly about the trade-off between accuracy, speed and how far a model can be pushed beyond what it has seen
• You have shipped models or simulations that survived contact with the real world, not just a validation split
• You are comfortable in front of customers. Site walks with operators, technical depth with researchers, plain language with practitioners, all in the same week
• You are uncertainty-aware. You build surrogates, you validate before you trust, and you are straight about where the model stops working
• Travel to customer sites does not put you off

Backgrounds that tend to work well: forward-deployed or solutions engineering roles with genuine ML depth behind them, or engineers out of infrastructure, energy or hard-tech environments who have already put ML or simulation into service somewhere physical.

Probably Not A Fit If

You are a pure research profile with no production or customer-facing track record, or a solutions engineer without the modelling depth, or you tend to defend a design rather than discuss it when someone pushes back.

The Practical Bits

• Location: Boston, hybrid and in person. Relocation supported for the right person
• Work authorisation: US citizens and green card holders only. No sponsorship, no transfers
• Compensation: Strong market base plus meaningful equity at an early stage, with competitive benefits. Happy to talk specifics early in the conversation
• Process: An initial technical screen, a take-home walked through live, a live coding session with senior engineers, and a final conversation with the founding team
• Note: This role carries US national-security and export-control requirements, which may affect eligibility for some candidates

• Applied ML Engineer, Physical Systems
• Boston (Hybrid) | Competitive base + meaningful early-stage equity