[…] is looking for an AI engineer who wants to commit to one product for the long run — someone who cares about the outcome, adapts quickly as the product evolves, and stays with a hard problem until users actually feel it is solved. Your mission will be to help a U.S. company in the tax incentives domain turn manual, expert-heavy workflows into reliable production software, where AI and ML extract, predict and suggest, and deterministic logic owns every final figure.
The product is live and in real use: the MVP is already working for users, and the team is actively building new capabilities on top of it. You will join a small core engineering team, work directly with the CTO, and see the impact of your decisions immediately. RESPONSIBILITIES As an AI Engineer (Typescript), you will be responsible for: Building multi-agent document pipelines that extract data from dozens of real client files — financial tables, long narratives, scans and diagrams — and hand it off reliably to downstream agents; Designing the boundary between probabilistic AI/ML output and the deterministic engine that owns every calculated figure, with validation, provenance and human review where needed; Integrating in-house ML predictions (scoring, estimation, anomaly detection) and defining what the product does when a model is uncertain, wrong or unavailable; Making AI behaviour measurable through tracing, evaluation and meaningful tests, so quality is observed rather than assumed; Building durable,