Interpretable Deep Learning
Reverse-engineering the circuits and representations inside transformers — turning black boxes into systems we can inspect, edit, and trust.
Open to research internships · Summer 2027
I build machine learning systems that areinterpretableandhonestly reliablein the open world.
Ph.D. Candidate @MIT CSAIL · Cambridge, MA
Reverse-engineering the circuits and representations inside transformers — turning black boxes into systems we can inspect, edit, and trust.
Making models reliable under distribution shift, with calibrated uncertainty and guarantees that survive contact with the real world.
Long-context attention and sparse computation — cutting the cost of foundation models without cutting corners on quality.
, Maya Zhang, Priya Patel, Wei Lin
, Sofia Ramirez, Jane Doe
, Jane Doe
, Wei Lin, Omar Hassan, Jane Doe
2023 — Present
Advised by Prof. Jane Doe · Trustworthy ML Group
2019 — 2023
Example University
Summa cum laude · Thesis on sparse attention
Summer 2026
Interpretability team — scaling circuit analysis to production-scale models.
Summer 2025
Conformal methods for open-world recognition.
2021 — 2023
Example Lab
First encounters with attention sparsification.
Say hello
The fastest way to reach me is email — I usually reply within 48 hours. Always happy to chat about research, internships, or collaboration.