Backend Engineer – FinOps & Agentic AI

About the Company

We need one engineer who can ship the full invoice-to-pay path (ingest, OCR, normalize, match, exceptions, ERP/payments integrations) and embed agentic AI on Azure: tool-using models, secure tool layers, RAG where needed, and human-in-the-loop for money and sensitive actions. You bridge domain workflows and modern LLM orchestrationnot two separate backend only and AI only roles.

About the Role

Invoice-to-pay (core pipeline)

• APIs for upload, storage (e.g. Blob), and multi-tenant persistence.
• OCR / document AI integration (e.g. Azure AI Document Intelligence), parsing into canonical invoice models, confidence and reprocessing.
• Normalization (SKU/UOM/pack) against master data; match engine rules (price, fees, duplicates, etc.) with Green/Yellow/Red outcomes.
• Workflow: exception queues, approvals, audit logging.
• Integrations: BILL.com (e.g. hold/release), Dynamics 365 BC (or similar) for AP/PO; secrets via Key Vault; idempotent, auditable calls.
• Async: queues and workers (e.g. Service Bus + Functions or workers) for heavy steps.
• Agentic AI (on the same platform)
• Azure OpenAI (or APIM to other models) with tool/function calling and multi-step flows.
• Tool layer: parameterized access to SQL, search, documentsnot unconstrained model access to data.
• Optional RAG (e.g. Azure AI Search) for contracts/policy docs with citations.
• Guardrails: Entra, tenant isolation, approval gates before side effects (payments, ERP posts, bulk updates).
• Observability and eval: tracing, logging, regression sets for agent behavior.

Responsibilities :

• Own end-to-end stories from upload through match to approval-ready state and payment/ERP hooks.
• Design and implement agent endpoints and shared contracts with the frontend (and solutions architect as applicable).
• Write tests for matching rules and critical integrations; contribute to OCR/match/agent quality gates with QA.
• Security by default: no secrets in code, parameterized queries, least privilege, clear audit trail.
• Document runbooks, failure modes, and reprocess paths.

Required qualifications

• 5+ years (or 3+ with very strong shipping experience) building production backend services.
• Strong Python (e.g. FastAPI) and SQL; multi-tenant and transactional thinking.
• Azure: compute (App Service / Container Apps), Blob, Key Vault, Service Bus (or equivalent messaging), managed identity.
• LLM integration: tool calling, multi-turn flows, not only single-shot prompts.
• REST/OAuth integrations with retries, idempotency, webhooks/polling.
• Testing discipline (pytest or similar) and code review habits that catch security and tenant bugs.

Nice to have

• Azure AI Document Intelligence, Azure OpenAI, Azure AI Search.
• BILL.com and Dynamics 365 BC (or similar ERP/OData).
• FinOps-adjacent domain: tolerances, 2/3-way match, compliance expectations.
• Evaluation harnesses for agents and OCR/match accuracy benchmarks.