JOB-1421
An AI startup that trains foundation models to help engineers design physical systems is hiring a software engineer to build the application, the agent layer above its models and the connective code between them. Expectations are high: each engineer is meant to deliver what would ordinarily need a handful of people, largely by working fluently with coding agents.
Responsibilities
- Own the customer-facing product end to end, covering the application, its APIs and the server side that exposes the models to practicing engineers
- Create the agent tier, including tool calling, orchestration, retrieval and feedback cycles that make raw model output dependable and editable for engineers
- Design and maintain evaluations for every feature you release, because unmeasured non-deterministic output is only a demo
- Handle CAD-adjacent engineering such as format conversion and geometry data wrangling
- Make infrastructure calls typical of an early company: how to deploy, what to monitor, and how to manage speed and spend
- Work with researchers and the product team, and define the in-house bar for working with coding agents
Profile
- Three to six years as a software engineer, with shipped software relied upon by actual users
- Daily, skilled use of coding agents, plus sound judgment about where they help and where they mislead
- Full-stack range spanning server code, APIs and customer-facing apps
- Real production work with LLMs and experience building evals
- Python, plus TypeScript or C++
- Comfort with ownership and ambiguity, and appetite for hard problems
- Pluses: CAD, 3D geometry or simulation exposure; mechanical or manufacturing background; early-startup product experience
Fully on-site in the San Francisco Bay Area, with a demanding schedule of 60 or more hours per week. Visa transfers and new sponsorships are supported.
