Job Description

Summary

Rud Pedersen is looking for a Technical Product Owner to own our secure internal AI platform based on LibreChat, and help turn the wider AI transformation programme into a practical, prioritised product roadmap.

This person will be the named owner for our internal platform: the backlog, roadmap, release priorities, user feedback, model/provider decisions, governance requirements, feature trade-offs, integrations, success measures, and handover into production support. The role is not pure project management and not a strategy-only role. It needs someone who can understand technical constraints, make product decisions, and work closely with engineers, IT, security, business users, and AI Champions.

This role is based in our Bulgaria office and you must have the existing right to work in Bulgaria. You will be working with colleagues across all our European offices.

What you will do

You will own the internal product roadmap and backlog. You will gather requirements from teams across Rud Pedersen, translate them into clear product priorities, define what should be built next, and make trade-offs between user value, technical effort, security, maintainability, and governance.

You will work with technical colleagues on LibreChat features, model and provider configuration, user permissions, file handling, prompt libraries, assistant patterns, RAG/knowledge retrieval, analytics, admin workflows, and integrations with Microsoft 365 and other internal systems.

You will define practical acceptance criteria for new features and workflows, test releases with real users, collect feedback, and decide whether something is ready to scale beyond a pilot. You will also maintain product documentation, release notes, guidance for users, and a clear view of known limitations.

You will be a bridge between business adoption and technical delivery. That means understanding use cases from public affairs, strategic communications, research, policy monitoring, workshop delivery, proposal support, and internal operations, then shaping them into realistic product increments rather than letting every request become “urgent”.