David Bau is an Assistant Professor of Computer Science at Northeastern University. My lab studies the structure and interpretation of deep networks.

We think that understanding the rich internal structure of deep networks is a grand and fundamental research question with many practical implications.

We aim to lay the groundwork for human-AI collaborative software engineering, where humans and machine-learned models both teach and learn from each other.

Alex Loftus is a first-year PhD student in David Bau's lab. Prior to interpretability research, Alex studied computational neuroscience, network statistics, and worked in machine learning for drug discovery. He is also a textbook author and Kaggle competition winner. He is primarily interested in applied interpretability projects in the code setting, as well as understanding knowledge acquisition during training.

Mechanistic interpretations of profound generalization capabilities in LLMs (such as instruction following, metalearning, or metareasoning)

Analysis or manipulation of the usefulness of chain of thought in reasoning models

Training (e.g., RL) methods for improving the faithfulness of generated explanations

Experiments that bridge the human-AI knowledge gap.

Interest in interpretability; reasonable familiarity with the current state of the field.

Demonstrated indicator of proficiency at doing sound research; doesn't necessarily have to be a paper, but some research output (blog post, exploratory jupyter notebooks, etc)

Strong programming fundamentals. Comfortable reading and writing code in python. Proficiency in the basic AI/ML stack.

The ability to own and execute a project; taking initiative to unblock technical challenges proactively when they arise, and plan new experiments without always being told exactly what to do.

Ability to understand a relevant body of scholarly work quickly (e.g., by assembling and reading relevant papers)

Ability to quickly implement a well-scoped idea and run a high volume of experiments.