The work

Each client arrives with its own languages, frameworks and older systems we have to learn quickly. You move between them, building interfaces, APIs and data models, and the AI parts inside them: agents that use tools, search over the client's own data, and the evaluations that show they work.

You use coding agents every day and you set them up properly: one harness per project, with its conventions, tests, permissions and context, so the code they write is code the team can trust. When an agent gets something wrong, you fix the harness as well as the code.

What you'll do

  • Build features end to end: interface, API, database, background jobs and deployment.
  • Design and ship AI features: agents that call tools, search over the client's data, streamed answers.
  • Set up and maintain the coding-agent harness for each project: instructions, tests, guardrails and review.
  • Write the evaluations that decide whether an AI feature is ready to release.
  • Work directly with the client's team when a decision needs what they know.

What you bring

  • Several years of shipping web products, and at least one you can show us running in production.
  • Fluency in at least two languages and their ecosystems, and the habit of learning a new one quickly.
  • Hands-on work with language-model APIs: prompts, tool calls, structured output, cost and latency.
  • Daily use of AI coding agents, and opinions about how to set them up.
  • Clear written English. Romanian helps but is not required.

What success looks like

  • It works in production for the people it was built for.
  • Another engineer can read it, change it and deploy it without you.
  • Its AI features are evaluated, and the results stay consistent after release.

Tools you'll use

  • TypeScript
  • React and Next.js
  • Python and FastAPI
  • Node.js
  • PostgreSQL and Supabase
  • Language-model APIs
  • Coding agents
  • Google Cloud