Senior Full-Stack Engineer (TypeScript/Python, LLM Applications)

About Us

Smirnov Labs is a fast-growing engineering company founded by an ex-Googler. We help product teams launch and scale quickly by leveraging modern technology and AI-assisted development.

Our projects are diverse — we partner with clients at the earliest stages and own the technical delivery end-to-end.

We're a small, senior-heavy team where every engineer has real impact. No bureaucracy, no hand-holding — just sharp people shipping fast.

This is an outstaff role: you'll be embedded directly with a client's engineering team as part of our delivery group.

The Role

Our client is a US healthtech startup whose AI platform helps dental patients understand, finance and book the treatment their dentist recommended. The product is live with dozens of practices and handles patient health data every day. The team is small, and the technical founder reviews every pull request.

The platform includes an AI treatment coordinator that answers patients' questions about their care, in text and voice. It's built and has been through several iterations, but it isn't in production yet — healthcare demands a higher bar. Your job is to get it there: build the evaluation harness, tune context and guardrails until the answers are safe and useful, then ship it. Next come agentic workflows, starting with booking appointments end-to-end. The agent runs in Python on Google ADK and Gemini; the product around it is TypeScript and Node.js, and you'll work across both.

What You'll Do

• Own the patient-facing AI assistant end-to-end — prompt and retrieval design, deployment, and monitoring in production.
• Build the eval harness: test sets from real patient questions, with automated scoring for safety, accuracy, resolution, number of turns, token cost and latency.
• Design guardrails for a healthcare setting: what the assistant may answer, what it must hand to a human, and how it treats sensitive data.
• Improve context passing — the patient's treatment plan, costs, insurance and history — so the model answers from facts, not guesses.
• Build agentic flows, starting with appointment booking against practice systems.
• Work across the stack: Python agent services, TypeScript/Node.js backend, React frontend, Postgres.
• Instrument LLM calls with logs, traces and metrics so regressions show up before patients see them.
• Work directly with the client's founder in a daily standup. No layers in between.

• Use AI-assisted development tools (Claude Code, Cursor, GitHub Copilot) as part of your daily workflow.

   

What We're Looking For

• 5+ years of production software engineering, including LLM-powered features shipped to real users
• Python and TypeScript/Node.js in production
• Hands-on evaluation work: you've built eval sets and harnesses and used them to make launch decisions
• RAG and context engineering: embeddings, retrieval, prompt design, tool use
• Guardrails and failure handling for LLM output — you know how these systems fail and how to catch it
• Production experience with at least one major LLM stack (Gemini/Google ADK, OpenAI, Anthropic), and the judgement to swap models when it pays
• Working Postgres and React — enough to ship a complete feature alone
• A track record of shipping without a detailed spec and without process handed to you
• Working US Eastern hours (required): daily 10:00 ET standup and availability through the client's working day
• Upper-Intermediate or higher English — you'll talk to a US founder every day

• Ability to own and drive features independently with minimal supervision

   

Nice to have

• LLM work in healthcare or another regulated domain (insurance, finance, legal)
• Voice agents: speech-to-text, text-to-speech, telephony
• Handling sensitive personal or health data in LLM pipelines: redaction, access control, audit trails
• Google Cloud: Cloud Run, Vertex AI

• Early-stage or agency background: greenfield work, shifting requirements, direct client contact

   

What We Offer

• Competitive salary above market average
• Fully remote work
• Flat structure — work directly with the client's founder and engineering team, no middle management
• Real technical ownership: you make the architecture decisions on the systems you build
• Diverse and technically challenging projects
• Modern AI-powered development workflow
• Small team culture — your voice matters, your code ships
• Paid vacation and sick leave
• Flexible schedule within the US timezone overlap
• Professional and career growth through real ownership, not courses and certificates