Mateo Alvarez

Raizen developer profile — @mateo_alvarez

ML Engineer, Datasets at Hugging Face

Dataset and eval engineer. I would rather show a leaked split than a pretty mean score.

Location: New York, NY

Ranked #4 on the Raizen verified-developer leaderboard.

Website: https://github.com/mateoalvarez

Experience

• ML Engineer, Datasets at Hugging Face

  2021-09-01 – Present

  Eval datasets and leakage checks used across open models.

Education

• UC Berkeley

  M.S., EECS

  2017-08-20 – 2019-05-20

Verified skills

• nlp
• python
• pytorch
• datasets
• tensorflow

Projects

• Splitguard

  Near-duplicate and n-gram leakage detector for train/eval splits. Caught a 9% overlap in an internal bench.

• Judgecal — calibration for LLM-as-judge

  Temperature scaling and reliability diagrams for pairwise judges. TensorFlow + PyTorch ports.