Domenico Lacavalla

Domenico Lacavalla

AI Researcher and Applied Scientist focused on NLP, computer vision, and LLMs. Building and deploying scalable AI solutions.

Data Scientist & Applied Scientist

Published in IBM, 2023

  • Implemented CPU multimodal retrieval (Grounding DINO, SBERT-Whisper) boosting speed 31% and recall to 83%.
  • Orchestrated ancient map geocoding (68.6% acc) fusing GenAI correction and DBSCAN under custom validation.
  • Bootstrapped a few-shot SBERT classifier achieving 73.4% accuracy via 10-shot LLM labeling; deployed via ONNX.
  • Saved $1,000+/mo on 5M+ samples via a hybrid LLM/SBERT anomaly detection pipeline for sentiment analysis.
  • Boosted summarization throughput 5x and cut waste 21.7% via parallelized LLM inference on A100 HPC.