Gerson C. Kroiz

Now

Senior ML/AI software engineer at Google working on optimizing AI training on TPUs and GPUs.

• Elastic Training and Resiliency on Google Cloud
• Efficient Checkpointing in Large Scale Jobs
• Talk at PyTorch 2025

Mechanistic Interpretability Enthusiast (MATS 9.0 Scholar w/ Neel Nanda)

• Why Did My Model Do That? Model Incrimination for Diagnosing LLM Misbehavior
• Principled Interpretability of Reward Hacking in Closed Frontier Models
• Can models be evaluation-aware without explicit evaluation?
More to come soon...

Prev

CTO of mindgrasp.ai

Research at ORNL, UCLA, NIST, and UMBC

Contact

Email / LinkedIn / X / GitHub / CV

Typeface: helvetica

Inspo: approaching shadow