Type: Full-time
Location: Remote — European time zones, plus at least 2 hours of evening overlap with the US (until ~20:00 Kyiv time)
Experience: Senior AI product engineering — seniority is measured by the quality of decisions and results, not years or titles
Track Record: Required — has taken LLM-powered features, ideally multi-agent or document-processing systems, from an ambiguous problem to production and supported them in real use
Engineering languages: TypeScript / Node.js; strong Python engineers are welcome if ready to work in TypeScript
Agreement Type: B2B / Contractor
Salary: Above market — fixed pay plus a KPI-based bonus
Reports to: CTO
Key Partners: Product lead, engineering team, client-side representative and tax subject-matter experts
About the Company
Our client is a long-established, independent specialty tax consulting firm — a national leader in tax credits and cost recovery, with offices across the U.S. and a team of 150–200 engineers, attorneys, CPAs and technologists.
The firm is turning decades of domain expertise into software: proprietary calculation engines, self-serve tools and integrations that let CPAs and property owners unlock six- and seven-figure tax incentives in minutes instead of months. It is a profitable, growing business, and the product is already in production with a multi-year horizon.
About the Role
You join a small, fully senior engineering team — up to five engineers, a product lead and the CTO — dedicated long-term to this one product. This is not staff augmentation or project rotation: the team owns implementation, technical decisions, delivery and production behavior.
The product combines LLMs and document intelligence with deterministic, auditable calculations. Much of the domain still runs on manual, expert-heavy workflows, so the work starts from ambiguous business problems rather than fixed specifications — with direct access to the subject-matter experts whose knowledge you are encoding.
The structure is horizontal: no direct reports, clear individual ownership of your area. Day to day you work with the CTO, the engineering team, the product lead and a client-side representative.
MUST-HAVE EXPERIENCE
AI in production, not just POCs: has shipped LLM-dependent or ML-enabled features into real use and supported them after launch.
Multi-agent or tool-calling systems: has built agents that read documents, extract data and hand results to other steps through clear data contracts.
Model failure modes: understands hallucinations, unstable output, quality drift and API unavailability, and can tell a concrete story of how an AI feature failed on real data and how it was contained without letting the system invent a value.
Evaluation and validation: measures the quality of LLM features with evaluation sets, tracing and regression checks, so quality is observed rather than assumed.
Confidence, fallbacks and human review: designs what the product does when a model is uncertain, wrong or unavailable, and knows where deterministic code must stay authoritative — in this product, every tax calculation.
Product engineering: takes a feature from domain model and typed APIs to a working React / Next.js UI and stable production behavior; startup or product-company experience is a strong plus.
Strong TypeScript / Node.js in production. Senior engineers with a strong Python background are welcome if ready to work in TypeScript.
Relational data: PostgreSQL schema design that outlived first assumptions, migrations on live data, auditability and history.
Asynchronous systems and integrations: queues, durable functions or workflow engines; retries, idempotency, partial failure and unreliable third-party services.
Testing and production readiness without a QA team: regression tests that catch real issues, debugging production failures, logging/tracing, CI/CD pipelines, safe releases and careful handling of sensitive data.
AI-assisted engineering: uses coding agents and LLM
