The company builds autonomous agents that run finance operations for its customers, covering accounts payable and receivable from end to end. The work spans invoice validation, dispute resolution, cross-company negotiation and cash optimization.
This is a senior individual-contributor role at the top of the engineering ladder. There are no direct reports, and you hold full technical authority over the architecture the rest of the team builds on.
Responsibilities
Design the multi-agent system: agent boundaries, event contracts, orchestration patterns, and where LLM reasoning stops and deterministic enforcement begins. You own that boundary for every action that moves money, plus the authorization gateway that keeps it fixed.
Build agents that read a claim, collect evidence from the purchase order, contract and delivery record, take a position and push it to settlement. They negotiate with the counterparty when suitable and hand off to a human when not. Deductions and short payments are classified, linked to the claim, and resolved rather than left to age.
Build agents that price early payment against the cash position on both sides, propose terms and execute them, so static payment terms become a continuously negotiated process.
Own the foundation of matching and reconciliation: three-way matching across purchase order, goods receipt and invoice with configurable tolerances; cash application against bank statements and remittance advices, covering partial payments and residuals; and the exception paths for every failure.
Build validation and anomaly detection: business rule checks, duplicate and near-duplicate detection, pricing discrepancy detection, and time-based anomaly detection. Precision matters as much as recall, since a false positive blocks a supplier and frustrates a procurement lead.
Design the integration framework, starting with ERP systems and then banking APIs, accounting platforms, Gmail, Drive and Slack. Cover the auth model, credential lifecycle and data normalization so that each new connector is a configuration change rather than a project.
Design compliance and security controls under a zero-trust model: independent verification of agent output against source data, review of outbound content, and protection against cascading failures from prompt injection. Also cover the European e-invoicing mandates arriving in France, Germany and Spain, and what EN 16931 requires of the pipeline.
Own the platform primitives every agent action depends on: an append-only PostgreSQL event store with tamper-evident hash chaining for replay and audit; cryptographic signature checks on agent events and approval chains before anything goes external; the Workflow Constraint Engine, which gates payments, emails and API calls on deterministic policy rather than model output; row-level multi-tenancy with tenant-scoped credentials in every data store; attribute-based authorization through Cerbos, applied identically to agents and humans; and the MCP security layer (auth, capability scoping, rate limiting) for external agent-to-agent access.
Set the engineering standard through architecture decision records, the review bar, on-call duty, and a clear record of what the team chooses not to build. Write the documents the next ten engineers will read in their first week, and spend time with controllers and AP leads watching agents fail in ways no test anticipated.
Requirements
8+ years of shipping production systems, currently at Staff, Principal, Lead IC or founding engineer level. You have carried an architecture through a scaling phase and lived with your own design decisions.
Strong systems design skills covering agent boundaries, failure modes, trust boundaries and contract evolution. You know money movement is not safe to retry by default.
Production depth in Python and/or TypeScript, and deep PostgreSQL knowledge: event sourcing, row-level security, indexing tradeoffs and schema evolution under load.
Hands-on work with LLM-based systems, agent orchestration frameworks (LangGraph, LangChain or similar) and document AI (LlamaIndex, LlamaParse or similar).
Clear, defensible views on when to use an LLM, a rules engine or traditional ML.
Experience designing data pipelines, integration architectures or event-driven systems in production.
Security awareness: you consider what happens when untrusted input reaches your pipeline, and you care about segregation of duties.
Technical leadership without a formal title: reviews that teach, designs that others adopt, and engineers who ramp faster because you are on the team.
Effective use of AI coding tools, with strong review skills for LLM integration and auth code.
Working knowledge across purchase-to-pay and order-to-cash. You need real depth in one of the three areas below and working fluency in the other two, enough to understand a controller's description of an exception before they finish.
Accounts payable: PO and non-PO invoice intake, GL and cost centre coding, approval routing and delegation of authority, three-way matching and tolerances, blocked-invoice and exception workflows, duplicate detection, payment runs, payment terms and early-payment discounts, supplier master data and bank-detail verification, and period-end accruals for uninvoiced receipts.
Accounts receivable: invoice issuance and e-invoicing, cash application and remittance matching, partial payments, short pays and residuals, credit memos, deductions and disputes, dunning and collections strategy, credit limits and credit risk, and aging and DSO.
Procurement: requisition to purchase order, supplier onboarding and qualification, catalogues and contract pricing, goods receipt, off-contract and maverick spend, sourcing and negotiation, and the handoff into AP where most problems arise.
Familiarity with the systems this work runs on: ERPs such as NetSuite, Dynamics or SMB-focused stacks, the sub-ledger to general-ledger relationship, and how period close limits what an agent may do and when.
Peppol and EN 16931 in production, or bank connectivity through ISO 20022.
Production use of cryptographic primitives: digital signatures, hash chaining and tamper-evident logs.
SOC 2, GDPR or data-residency engineering in a regulated environment.
Founding experience, or time as one of the first five employees at a startup.
Conditions
Individual-contributor role with no direct reports.