Mateusz Opala

Led architecture and delivery of core agentic products (Deal Inspection Agent, Ask Clari, automated email generation), reducing manual deal review from hours to minutes and established evaluation and release gating enabling multiple safe releases per day.

Improved retrieval precision from ~40% to ~70% using neural reranking (RoBERTa; Mistral 7B via QLoRA) and replaced GPT-4 relevance scoring with in-house models to cut inference costs by orders of magnitude.

Led LLM-powered features for a global Q&A platform serving 350M+ MAU and designed hybrid AI systems (retrieval + neural + symbolic) that increased answer coverage from ~50% to ~80%.

Founded and scaled the ML department from 0 to 8 engineers, defined ML project discovery and delivery processes, and built a revenue-generating ML practice achieving seven-figure annual revenue.

Co-founded and served as CTO, led ML architecture and a team of 6 engineers to design and ship an ML-powered visual search and image-to-catalog matching system for fashion, enabling relevant product ads.

Implemented real-time CNN inference pipelines for medical imaging workflows, contributing to production-ready imaging solutions.

Completed a Bachelor of Computer Science with a thesis on deep learning for semantic image search, focusing on applying CNNs to image retrieval problems.