About Position

Location: NYC, NY

Our client is building the AI operating system for the CFO office, connecting to a company’s existing finance stack and automating the work behind the close, revenue recognition, reporting, cash management, and audit readiness. The goal is not to help accountants write better prompts but for finance teams to review exceptions while AI does the rest, with audit-ready agents that can reason, explain their decisions, escalate uncertainty, and continuously improve. The company has raised $9 million led by Foundation Capital, alongside Aravind Srinivas (CEO of Perplexity) and finance leaders from Ramp, Gusto, Zuora, and the Big Four.

About the Role

We are hiring a Staff AI Engineer to own the platform layer that every engineering pod builds on, from agent frameworks and context systems to orchestration, verification, observability, and data infrastructure. This is not a module-ownership role: your work multiplies the output of the entire engineering team, and you identify the bottlenecks limiting the company rather than waiting for specs from a PM or an architecture committee. With 2 or more years of direct agentic AI work and a research-oriented ML or NLP background, you bring the depth to define the abstractions that make hard problems simple across a lean, high-ownership team in New York City.

What You’ll Own

• Own the agent harness the entire company builds on, defining abstractions for context, verification, guardrails, observability, and developer tooling so every pod ships audit-grade AI agents on shared rails.
• Design and build the financial context graph, the structured layer every agent reasons over, including ledger state, policies, contracts, precedent, and entitlements, kept coherent, scalable, and multi-tenant safe.
• Define the verification, auditability, evals, and observability standards that determine whether AI output is safe enough for a customer’s books, and enforce them by construction.
• Own the ingestion, normalization, reconciliation, and canonical ledger model that turns data from ERPs, banks, billing, payroll, CRMs, and email into a trustworthy source of truth.
• Design the durable execution layer for long-running AI workflows and the exactly-once, audit-ready path that safely writes back to ERPs and systems of record.
• Build architectural primitives and frameworks that become the foundation other engineers depend on, compounding leverage across the full engineering org over time.

Must-Have

1. 2 or more years of hands-on agentic AI work, with the ability to speak deeply about agent architecture, context systems, orchestration, and technical tradeoffs.
2. 8 or more years of overall software engineering experience, with deep expertise in distributed systems, data infrastructure, workflow engines, transactional systems, integration platforms, or AI and agent platforms.
3. Research-oriented background in ML, NLP, or AI agents, with demonstrated ability to apply that depth to production systems rather than purely academic work.
4. Strong backend proficiency in a modern language, with Python as the primary stack; candidates with deep experience elsewhere must be prepared to ramp fast.
5. Track record of building platform systems that multiplied the output of other engineers, including designing abstractions or infrastructure that multiple teams depended on.
6. Experience thriving in an early-stage startup environment (pre-seed through Series B), where ownership is high, resources are constrained, and there is no separation between strategy and execution.
7. In-person at the New York City office. (Startup hours, 6 days a week, 9 am to 7 pm or 8 pm)

Nice-to-Have

• Prior Staff or Principal IC experience at a high-growth startup, having built systems that multiple teams depended on.
• Background in accounting, fintech, ERP, payments, treasury, audit, or compliance software, or experience building ledger and sub-ledger systems.
• Experience designing and operating AI agent platforms in production, including context systems, evals, verification, guardrails, observability, and orchestration.
• Experience building systems under SOX, SOC 1/2, financial audit, or other regulatory requirements where correctness mattered more than speed.
• Fluent use of AI coding agents (Claude, Cursor, or similar) as a daily collaborative tool for shipping significant production systems.