Lorenzo Germini

• Complaion helps European SMEs implement, maintain, and get real value from ISO certifications.
• Building the AI engine behind it: agentic RAG over ISO documentation, and the evaluation frameworks that show when it can't be trusted in production.
• Shipping AI product features and automation workflows end-to-end, from prototype to production.

• Built the backend and GenAI API layer from zero on AWS (DynamoDB, Amplify, Lambda) for an EdTech product.
• RAG stack on pgai with custom parsing pipelines for machine-readable and handwritten content, plus tracing and evaluation workflows to tune prompts and retrieval quality.
• Led MVP engineering, managing two other developers.

• Used Python and multi-source sensor data to find inefficiencies across pharmaceutical utilities; designed water rejection strategies projected to save 10,000 m³ annually.
• Anomaly detection and time-series forecasting for manufacturing operations, plus semantic NLP for fault detection and maintenance decisions.
• Natural-language interfaces and generative AI use cases that made plant data usable on the floor.

• Taught math and physics at my old high school; built an interior-design MVP with Stable Diffusion and Dreambooth fine-tuning on a hand-curated dataset of room designs I preprocessed myself, then killed it on the evidence after customer discovery.

• Fine-tuned language models for abstractive summarization of biomedical literature, deployed as an end-to-end pipeline for the company's first digital health product.

• Took a personalized-medicine filling prototype through hardware-software integration testing, and built a gradient-boosted tree ML model for fill parameter tuning, 20% faster than manual.

AI game companion for La Ghigliottina, the cult final round of Italy's most-watched quiz show. Send the five clue words, typed or snapped from the TV, and it finds the one that connects them. 100% accuracy so far, with rate-limit handling because Italians take this game seriously.

MSc thesis work: an end-to-end pipeline that turns full-text biomedical papers into either a detailed analysis or a TL;DR, tuned on longevity and human-health literature.