Senior Software Engineer, Expense

Navan (Nasdaq: NAVN) is the AI-powered business travel payments and expense platform that makes it easy to book pay and close the books. Travelers get a product they actually want to use. Finance teams get real-time control policy and reconciliation. Expense is a core part of that story: from the moment a receipt is captured to the moment it lands in the general ledger.

About the role:

We are growing Expense Engineering in Berlin and London. You will join the org that owns the full expense lifecycle for employees approvers and finance admins used by companies worldwide including complex EMEA tax VAT per diem mileage and multi-entity setups.

This is not a research-lab role and not a generic chatbot posting. You will ship in the same systems that already run receipt intelligence conversational expense capture missing-info assistants and admin automation then raise the bar on accuracy latency cost and safety.

You may land on employee experience (web/mobile expense) international spend (itemization per diem mileage tax receipts) admin and control (policy custom fields audit) or adjacent platform work (cards statements ERP). The product problems and the AI problems are shared.

What youll work on:

The expense product end to end

• Employee flows: transaction entry drafts receipts trips reimbursements cards and bank-linked spend
• International spend: line-itemization per diem mileage VAT/GST and tax receipts
• Admin and control: policy approvals custom and HR fields audit logs reporting
• Money movement and close: statements purchase cards reimbursements accounting postings and inbound company/user data

AI that is already in the product

• Receipt intelligence: extract merchant amount date tax and line items from receipts and documents at high daily volume. Improve quality latency and cost; compare and fail over between extraction providers including large language models where they outperform traditional OCR.
• Conversational expense: multi-agent assistants that help people create and complete expenses in natural language (mileage policy personal context knowledge retrieval). Streaming experiences on web and mobile and the same command path in workplace chat tools.
• Missing-info assistants: agents that find required fields still empty (cost objects GL project department description) ground suggestions in the users history and trips stream progress and write back through existing expense APIs with safe partial updates clear provenance and guardrails against prompt injection and off-scope use.
• Admin and finance agents: assistants that can approve reject flag route and explain spend using tools over our APIs search and warehouse data without bypassing permissions or audit.
• Quality cost and safety: evaluation harnesses for accuracy and latency; kill switches; careful handling of PII and payment-sensitive data on receipts and transactions. Prefer deterministic code for policy catalogs and tax rules; use models for language ranking and tool orchestration.

What youll do:

• Design and ship features in backend services (Java and Spring Boot) and/or web (TypeScript) with AI services in Python or TypeScript where that is the right fit
• Own APIs and event-driven flows that move money receipts and policy decisions at company scale
• Build or extend LLM-backed agents: tool use grounding streaming session memory retrieval and evals not one-off prompt demos
• Improve document extraction and auto-itemization: quality tax fields latency cost and provider failover
• Instrument production: observability quality dashboards chat satisfaction extraction accuracy vs. human correction
• Partner with Product and Design on employee admin and mobile AI surfaces (including unified travel-and-expense chat)
• Review for correctness in financial systems: idempotency auditability permissions no silent partial writes
• Mentor others and write designs that teams across time zones can implement from

What were looking for:

• 5 years building production software (8 for Staff). Strong CS fundamentals.
• Depth in at least one of: Java/Spring Boot modern TypeScript or Python services. Willingness to cross the boundary.
• Experience with relational data modeling APIs and cloud (AWS). Messaging systems (Kafka or similar) are a plus.
• Shipped AI in a real product not only notebooks: LLM APIs tool/function calling retrieval-augmented generation evaluation or document/vision extraction. You can talk about failure modes (hallucination cost latency prompt injection PII).
• Comfort with financial or workflow-heavy domains: expenses payments ERP tax or similarly high-stakes data.
• Clear written design strong code review and bias to production quality (tests CI/CD observability incident ownership).
• Based in Berlin or London (or willing to relocate). Collaboration across US Israel and India time zones.

Nice to have:

• document OCR; VAT/GST or multi-currency; policy engines; agent frameworks; eval platforms; mobile AI UX.

Why this team why these cities

Expense is where Navans AI-powered T&E claim has to be true: receipts chat and admin agents sit on the same ledger as reimbursements and accounting close. Berlin and London put you next to EMEA customers who care about VAT per diem and audit and next to a growing Expense engineering presence not a satellite office doing overflow tickets.