Embed with customers to map out workflows, de‑risk constraints, and define crisp success metrics for custom builds.
Design → prototype → productionize custom workflows: customize knowledge sources and retrieval pipelines, tool use/agents, prompts, and guardrails; then harden them for reliability, observability, and scale.
Integrate workflows with clients systems (DMS, KM, ticketing, identity/SSO) and data sources; standing up secure connectors to the rest of the Harvey platform.
Build and maintain evals & harnesses that capture real‑world quality on a client-by-client basis, wiring those signals into iteration loops and model choices.
Operationalize adoption: run training, write crisp runbooks, and hand-off durable playbooks to customer champions—and to Harvey product/eng—so wins scale beyond one account.
Surface field patterns (recurring prompts, tools, workflows, failure modes) that inform platform capabilities and future product bets.
Requirements
2+ years building and operating production software with meaningful 0→1 ownership and the ability to operate under ambiguity.
Experience building LLM‑powered applications (retrieval, tools/agents, structured outputs, prompt/runtime safety) and taking them to production.
Comfort working directly with customers—from scoping ambiguous problems to shipping, integrating, training, and iterating on adoption.
Practical experience with evals (designing task suites, pipelines, and dashboards that reflect user quality); you use evals to drive model/product decisions.
Clear, concise communication; low‑ego collaboration; appetite for in‑person pairing with teammates and customers in NYC.
Benefits
Comprehensive health, dental and vision coverage
Retirement benefits (401k match up to 4%)
Flexible PTO
Applicant Tracking System Keywords
Tip: use these terms in your resume and cover letter to boost ATS matches.
Hard skills
production softwareLLM-powered applicationsretrievalprompt safetytask suite designpipeline designdashboard creationmodel decisionscustom workflowsdata integration