About the role
Responsibilities:
- Design and ship full-stack AI features end-to-end, from the user-facing surface to the services and agent runtimes behind them.
- Build and operate agent harnesses (tool use, planning, memory, eval loops, guardrails) that reason over long, complex contracts and the varied document types around them.
- Own retrieval and context systems for high volumes of complex documents: chunking, embeddings, vector indexes, hybrid search, re-ranking, and citation-grounded generation.
- Take prototypes to production: observability, evaluation, regression suites, latency and cost budgets, prompt and model version management.
- Drive system-design decisions across services, data stores, model providers, and infrastructure.
- Use coding agents fluently as a daily driver to ship more, faster, without losing quality.
Requirements:
- 4+ years of professional software engineering experience, with 1-2+ years shipping generative AI applications to production.
- Strong front-end (Next.js / TypeScript) and Python back-end skills, with a solid grasp of system design.
- Working knowledge of LLM fundamentals and a clear-eyed view of their limitations and failure modes.
- Has shipped at least one RAG system or agent to production, and built the agent harness around it (orchestration, tools, evals, observability).
- Hands-on experience with an agent framework (LangChain, LangGraph, LlamaIndex, or similar), or has built their own agent runtime/harness and can clearly justify why it beats reaching for a framework.
- Production experience with a major cloud (AWS / GCP / Azure), a relational database (Postgres preferred), and object storage.
- Fluent daily use of coding agents (Claude Code, Cursor, Copilot, or similar).
Even Better If:
- Experience with model evaluation tooling (Braintrust, LangSmith, Arize, Phoenix, LangFuse).
- Background in a compliance-heavy or regulated domain, working with sensitive data (SOC 2, ISO 27001, HIPAA-style regimes) and the engineering practices that come with it.
- Open-source contributions of any kind.