Matteo Giomi

Directed R&D initiatives focusing on anonymization and text data protection for LLMs and AI systems. Designed and deployed generative AI solutions integrating privacy-enhancing technologies, unlocking new LLM use cases. Improved Named Entity Recognition (NER) pipelines, boosting sensitive data detection accuracy up to 97% on av erage and reducing latency to sub 50ms, enabling real-time protection in production environments. Integrated and deployed the product into other AI systems, like NVIDIA Guardrails.

Co-engineered a real-time sensitive-data protection layer for enterprise LLM and search pipelines, en suring data protection without sacrificing performance. Built and delivered engineering demos interfacing with multiple AI frameworks. Developed core product and engineering demos interfacing with several AI frameworks.

Provided expert guidance for the NIST Collaborative Research Cycle (CRC), shaping evaluation method ologies for de-identification and synthetic data privacy risk. Designed and executed large-scale privacy analyses across 300+ datasets providing a solid benchmark to evaluate data anonymization solutions from industry and academia.