Hi! My name is Stan, and I want to create machine learning models that generalize.

I am particularly interested in methods that improve generalization when labeled data is scarce and noisy. In the past, I’ve explored:

1. Large domain-specific pre-training datasets for transfer learning with microscopy images
2. Supervised contrastive pre-training to improve generalization of ultrasound video models across hospitals
3. Large language models for automated soft-labeling of domain-specific text with prompt engineering

I am also a believer of slow science when possible.

Shoot me an email if you’d like to chat! And do include the word “stupefy” in your email.