About the Role

This is a hands-on ML engineering role focused on building and shipping intelligent agent systems that automate complex accounting workflows. You will own projects end-to-end, from scoping and experimentation through to production, and operate with a high degree of autonomy on a small, high-caliber team. The work sits at the intersection of applied research and product engineering, with direct impact on how AI transforms a major professional services industry.

What You'll Do

Design, build, and iterate on multi-agent architectures that automate real accounting workflows end-to-end.
Define and enforce autonomy boundaries, tool usage, and fallback behaviors that keep agents safe and reliable.
Manage context and memory across multi-step agent loops, routing and optimizing models under latency, cost, and accuracy constraints.
Build scalable evaluation pipelines (offline and online) that run experiments automatically and surface regressions early.
Define golden tasks, labeling strategies, and metrics that make model performance measurable and comparable over time.
Architect prompt stacks, instruction hierarchies, and retrieval pipelines that structure and surface relevant context for model reasoning.
Parse unstructured documents into structured representations and design guardrails and validation layers to keep behavior deterministic.
Scope projects clearly, write concise specs and architecture docs, and communicate progress to your pod regularly.

What We're Looking For

2 or more years of hands-on ML engineering experience at a fast-paced startup (Series A to D), a top-tier tech company, or a hedge fund or AI-native organization.
Proven track record building end-to-end LLM-based AI agent applications, including model orchestration, benchmarking, and evaluation frameworks.
Deep expertise in Python and transformer-based or LLM-based systems.
Experience running structured ML experiments: framing hypotheses, building evaluation infrastructure, and iterating on measurable results.
Strong fundamentals in machine learning, with a hands-on emphasis on agent systems and evals rather than classical ML alone.
Technical degree (CS, Physics, Math, or similar) from a strong academic program.
Clear, concise communicator who can break complex concepts down to their core elements.
Currently based in the US or Canada, with preference for New York or the East Coast.
Background or genuine interest in accounting, finance, or economics is a plus.