At Prometeia S.p.A., I own production infrastructure for a multi-tenant computer vision service that estimates car damage. I expanded image-processing capacity and cut heavy-claim latency by about 70%, while optimizing Triton GPU inference to reduce p95 latency by about 30%.
I’ve also built vector search for fraud detection and release verification tooling at Prometeia. Previously, at REPLY S.p.A., I developed data platform pipelines and extended production ML pipelines on AWS; I also built and published askai, an open-source Python CLI for LLM queries.

