About this role
Overview: You will join a newly created 5-person squad responsible for maintaining and evolving a portfolio of production AI agents serving the back office of the securities services (custody) division of a major European bank. The agents are built in Python on CrewAI, with a planned migration to an internal SDK and LangGraph. Surrounding APIs are written in Go. Agents are containerized (Docker) and deployed on Kubernetes; the team owns runtime configuration through ConfigMaps. What will you do? Design, build, and continuously improve the AI agents themselves: their reasoning flows, tool use, prompts, and evaluation. Adapt central general-purpose agents (drafting, summarization, process driving) to custody back-office use cases. Develop and maintain agents in Python using LangGraph, LangChain and the internal SDK. Design agent workflows: task decomposition, tool/function calling against APIs, memory and state handling, guardrails, and human-in-the-loop checkpoints appropriate to back-office controls. Engineer, test, and version prompts; manage prompt/config changes through Kubernetes ConfigMaps with proper release discipline. Build and run agent evaluation: golden datasets from real back-office cases, regression suites, quality metrics (accuracy, groundedness, escalation rate), and shadow testing before rollout. Localize central agents: analyze the gap between general-purpose behavior and local business requirements (formats, vocabulary, workflows, controls) and implement the adaptation layer. Instrument agents for observability: structured logging of reasoning traces, token usage, latency, and failure modes. Handle model lifecycle concerns: model/provider changes, context window constraints, cost/latency trade-offs. Work daily with the Business Analyst to translate operational knowledge (settlements, corporate actions, client queries, reconciliations) into agent behavior. What are we looking for? 3–6 years of Python engineering, with at least 1–2 years on LLM/GenAI applications. P ractical experience with at least one agent framework (CrewAI, LangGraph, LangChain, AutoGen, or similar); understanding of ReAct-style loops, tool calling, and structured outputs. S olid grasp of prompt engineering, RAG patterns, and LLM evaluation techniques. Comfortable consuming REST/gRPC APIs (Go backend); JSON schema design; async Python. Working knowledge of Docker and Kubernetes basics (enough to deploy, read logs, and edit ConfigMaps safely). Testing culture: pytest, mocking LLM calls, deterministic test design for non-deterministic systems. Fluent in English What can you expect from us? A permanent job contract for a long term project; Tech equipment + SIM Card + personal smartphone; Health and Life Insurance; Social events and team buildings; The commitment of letting you grow with us, and be rewarded accordingly; A dynamic and young team that will be always there to support you; Training in the latest technologies; Coffee, fruits, snacks and a warm welcoming when you pass by the office.