Job Description

This is a remote position. A global consumer goods brand runs a customer-facing AI agent - agentic RAG over its product range - that answers customer questions in natural language. It is already live in couple European markets and is onboarding roughly one new country every month, each with its own language and regulatory nuances. The programme is sponsored at board level and carries high internal visibility.This role exists to close the gap between functional design and buildable design. You take the functional (non-technical) design, resolve as many of the unknowns as you can, and produce a low-level design the development team can build from without rediscovering the same problems. In practice that means writing real prototype code, proving integration patterns end to end, and documenting what works clearly enough for a distributed team to implement it. WHAT YOU'LL BE DOING Turning functional designs into low-level technical designs that development teams can build from directly Running spikes and proofs of concept independently - scoping them, building them and reaching a clear conclusion Proving integration patterns to surrounding systems end to end: authentication, authorisation, secrets management and API contracts Designing and prototyping agent behaviour with LangChain and LangGraph; instrumenting, tracing and evaluating with LangSmith Implementing MCP integrations and agent-to-agent communication patterns Writing documentation and reference code clear enough for a distributed team to follow without you in the room Working alongside the tech lead and architects so that prototype decisions stay consistent with the target architecture Driving open questions to resolution across teams - raising the issue, convening the right people and closing the loop Requirements Hands-on experience with LangChain, LangGraph and LangSmith — building, debugging and evaluating agent systems, not just familiarity Strong, current Python development skills Experience building interfaces to other systems, including authentication, authorisation and secrets management Practical experience implementing Model Context Protocol (MCP) integrations Experience with agent-to-agent (A2A) integration patterns Able to conduct proofs of concept independently, from scoping through to a documented conclusion Comfortable working closely with both a development team and solution architects Proactive communicator — able to set up meetings, chase answers across teams and resolve blockers without being asked Clear technical writing and example code that others can work from NICE TO HAVE Experience with telephony and call centre applications - the platform uses Twilio Experience with multilingual or multi-market AI deployments RAG evaluation and answer-quality measurement at production scale Experience working with distributed teams across time zones Exposure to regulated or compliance-sensitive content domains WHAT SUCCESS LOOKS LIKE Development teams start work with fewer open questions and fewer false starts Integration patterns are proven once and reused across markets, rather than rediscovered each time Low-level designs are specific enough to build from - and are actually used by the teams they were written for Technical unknowns surface and get resolved before they become delivery blockers TECH ENVIRONMENT Core stack Python LangChain, LangGraph, LangSmith Agentic RAG over product content; multi-language and multi-market MCP and agent-to-agent integration Authentication, authorisation and secrets management across system boundaries Twilio for telephony and call centre channels Working context Product owner, management and tech lead based in Poland English as the working language across the delivery team Benefits