AI Automation Engineer İş İlanı

AI Automation Engineer (Python) Employment Type: Full-time, Remote – Türkiye Experience: 4–8 years Industry: AI Automation / Fintech About the Role We are looking for an AI Automation Engineer (Python) to join RDC Partner. In this role, you will work with the engineering team of our US-based business partner in the fintech industry. Our business partner is developing an AI-powered agent platform designed to automate key business workflows across onboarding, underwriting, risk management, and customer support. In this role, you will be responsible for building and enhancing AI agent systems, developing backend services and integrations, and ensuring the reliability, security, and scalability of AI-driven workflows. You will work remotely with teams based in the US and Türkiye, contributing to the development of secure, reliable, and production-ready AI automation solutions. Responsibilities Build and extend the agent control plane: FastAPI services, Pydantic v2 models, SQLAlchemy 2.0 over PostgreSQL. Implement agent rounds — the coordinating agent and its domain sub-agents — with explicit retry budgets, confidence handling, and escalation to human review when the system should not decide on its own. Keep calculation and policy in code rather than in the model, and keep a scripted policy path working as the reference implementation and fallback. Build and maintain MCP servers that give agents governed access to external systems — the processor interface, Microsoft 365 mail and Teams, Zoho, internal document stores, web search — read-only by contract, with a single audited send path. Enforce the outward data contract across every connector: redaction, reference tokens, receipts, and a consistent item shape. Work on the document pipeline: offline OCR of scanned packets, person-name redaction and PII masking before any content reaches an external model. Build idempotent write paths with receipts, and a shadow-versus-live switch per agent so a new agent can run alongside the existing process before it is trusted with it. Handle encryption of identity fields, fingerprinting and key rotation. Instrument the system with OpenTelemetry so every agent round, tool step and send is a span you can inspect afterwards. Build background work on the thread-based scheduler, using database advisory locks for per-merchant serialization. Contribute to the React and TypeScript operator panel — starting and stopping agents, interacting with them, and showing metrics. Maintain and extend the pytest suite, which is the main thing standing between a prompt change and a bad decision in production. What we are looking for 4–8 years building production services in Python, with recent hands-on work in a modern async stack. FastAPI or a comparable ASGI framework, Pydantic v2, SQLAlchemy 2.0 and PostgreSQL in production — not just in a side project. Real LLM application experience: tool and function calling, structured output, prompt design, and handling non-determinism in a system that has to behave predictably. A clear view of what belongs in a model and what belongs in code, and the discipline to keep the line in the right place. Integration experience with third-party business APIs — authentication, rate limits, pagination, partial failure, and the ways vendor APIs lie to you. Idempotency, receipts and audit trails: you have built a write path that could be retried safely. Comfortable with tracing and observability, and in the habit of instrumenting rather than adding print statements. Strong testing discipline. A large test suite is part of the job rather than an afterthought. A security mindset around personal data — you treat a redaction gap as a serious defect, not a cosmetic one. Daily hands-on use of AI coding tools (Claude Code, Cursor, GitHub Copilot or similar), with a clear and specific view of where they help and where they mislead. Hands-on and comfortable in a deliberately minimal stack, as described above. Strong written and spoken English. You will be writing for and speaking with colleagues in the United States every week.