Alex Lu

Senior Researcher

Microsoft Research


My name is Alex, and I develop machine learning methods for biology. High-throughput methods generate an unprecedented amount of data. My research focuses on machine learning methods that help us discover hypotheses from millions of microscopy images or protein sequences:

  • Many methods require millions of manually-labeled images as training data. This process is time-consuming and introduces bias into our methods:
    I create unsupervised methods that do not require extensive expert labor.
  • Robustness: we want models to identify biology consistently even when there are technical differences, like cells being illuminated differently in images.
  • I research ways to use machine learning to discover novel biology: unsupervised and self-supervised methods can operate as independent experts that may be sensitive to different things than human experts.


My research focuses on making machine learning practical for biologists.

Out-of-Sample Generalization

How do we build models that are robust to small differences in future experiments?

Representation Learning

Self-supervised deep learning produces unbiased representations of biology

Exploratory Biology

Using big image data to discover new biological functions for proteins

Recent Publications

Improved Conditional Flow Models for Molecule to Image Synthesis