Domenico Lacavalla

Domenico Lacavalla

Data Scientist and ML Engineer focused on applied AI. I build and deploy solutions based on NLP, computer vision, and LLMs.

Data Scientist AI Associate

Published in IBM, 2023

  • Utilized generative AI and fine-tuned BERT to analyze 4M conversations, identify trends, apply clustering, and present results to stakeholders.
  • Built and deployed customer segmentation models handling 6M data points, integrating a full training pipeline with automated retraining.
  • Engineered production AI pipeline leveraging OCR/OMR, NER, LLMs and Speech-to-Text for 10M+ assets, optimizing digitization to minute-scale.
  • Applied ResNet (86% score) & CLIP (93% score) to evaluate fidelity of OMR output (Audiveris) against original digitized scores.
  • Initiated HPC (H100 GPUs) exploration and parallelization strategy using SLURM aiming to halve processing time for core digitization primitives.
  • Optimized complex ETL query execution from 20 hours to 1 hour by refactoring SQL and leveraging Spark on IBM Cloud with multithreading.