Applied ML Engineer, Agents (AI for Accounting)

An AI startup building autonomous agents that perform real accounting work is looking for applied machine learning engineers. Each hire takes full responsibility for projects that put intelligence into production, acting as both researcher and builder: planning the work, setting success measures, running experiments, and deciding when a system is ready to ship.

What you’ll own

• Design and refine multi-agent architectures that automate genuine accounting workflows, with clear limits on autonomy, tool use and fallback behavior
• Manage context and memory so agents stay coherent over many steps, and route and tune models against latency, cost and accuracy constraints
• Create evaluation and experimentation pipelines, offline and online, with golden tasks, labeling approaches and metrics that make results comparable
• Instrument systems to catch regressions, track error categories and drive continuous improvement
• Engineer prompts, retrieval and indexing, and document parsing so agents reason over messy inputs, with guardrails and validation layers
• Write concise specs, report progress plainly, and share learnings with your pod

Qualifications

• Four to twelve years as a machine learning engineer, including startup, quant or top-tier tech experience
• Hands-on work with LLM-based agent applications, including benchmarking, orchestration and behavior evals
• Track record of structured experiments: framing hypotheses, building eval infrastructure, iterating on measured results
• Strong Python and a clear, first-principles communication style
• Applied rather than research orientation
• Comfort in a startup office five days a week

On-site in New York, NY. Visa transfers considered.