Maciej Tonak

At Capgemini, I design and ship AI-driven solutions in Python for enterprise clients, from prompt and context design through structured outputs, model selection, and ground-truth evaluation.

I'm currently designing an extraction pipeline for an enterprise invoice-processing platform handling up to 20M documents annually. I own extraction quality, benchmark OCR engines and multimodal models against labelled documents, and help maintain 97% field-level extraction accuracy and 80% straight-through processing.

I also built an embeddings-based recommendation system for routing P1/P2 support tickets, reducing mean assignment time by about 60%. My work includes SAP validation logic, APIs, databases, containerised delivery pipelines, and GenAI automations shaped directly with business stakeholders.

Alongside delivery work, I've run client requirements gathering and technical training. My experience as a science educator also helps me explain technical material clearly to non-technical audiences.