From the business case to the build: I scope AI use cases, design the evaluation that proves they work, and ship production AI apps on AWS, including multi-agent systems, LLM evals and MCP servers. 11+ years in data and AI, now a tech PM at Amazon in Luxembourg.
Live AI applications I designed, built and deployed through T4Tangerine, from multi-agent systems on AWS to MCP servers and open datasets.
Shows a company where it overpays for technology and what to change, tool by tool, shaped by its own policies. An orchestrator agent on Claude calls a RAG agent over the client's documents and a pricing catalogue; every cost and saving is calculated in code, so each figure can be audited.
Turns a goal into sub-goals with dependencies, the roles and capacity needed, hiring gaps, costs, the P&L line it moves, and risks. A signed-in app on AWS with Claude on Amazon Bedrock.
Asks Claude, ChatGPT, Gemini and Perplexity the questions buyers ask, then measures whether a brand is named, where it ranks, which competitors appear instead and which sources are cited. Runs as a signed-in app with reports across all four assistants.
A source-linked tracker of fines, court rulings, regulator orders and guidance under GDPR, the AI Act, DORA and the Digital Services Act, tagged for financial services. New entries are drafted by Claude against the taxonomy and reviewed by a person.
A remote MCP server that lets Claude, or any MCP client, turn an Instagram data export into organised albums in Google Drive or Cloudinary: it recovers capture time, captions and GPS from metadata, groups media into albums and uploads them, safely re-runnable from a checkpoint.
Most AI projects don't fail on the model. They fail between the people who want the outcome and the people who build the system. That gap is where I work.
I started in data science, building machine learning models and data pipelines at INSOFE, Optum and Target. At Amazon I moved into program management, and today I take generative AI agents from a leader's idea to launch: scoping the use case, writing the value case, building the evaluation that proves an agent is ready for production, and driving adoption once it ships.
Outside work I build. Through T4Tangerine I design and deploy multi-agent applications on AWS, Claude API tools and MCP servers for European businesses, and I'm studying for an M.Sc. in Artificial Intelligence at the University of Surrey.
Delivering for business stakeholders across HR, retail and healthcare, from machine learning models and data pipelines to generative AI agents.
Long-form posts on AI agents, reliability and governance, published on T4Tangerine. See all posts →
Open to conversations about deploying generative AI in the enterprise: forward-deployed and solutions roles, speaking, and AI projects in Europe.
