Michelle Pan

Michelle Pan

Berkeley, CA, she/her


2024 — Now
M.S. Computer Science at Stanford University
2020 — 2024
B.A. Computer Science at University of California, Berkeley

AI, logic, and language.

Work Experience

2022 — Now

Investigating use of reinforcement learning for building brain co-processors to restore injured neural circuits, under Dr. Erdem Bıyık and Prof. Anca Dragan.

2022 — 2024

Taught discussion sections, holding office hours, and grading exams for CS 70: Discrete Mathematics and Probability Theory for 5 semesters.

2022 — 2022

Worked on the Perception team at an Alphabet robotics moonshot, collaborating with researchers at Google Brain. Integrated vision-language models for zero-shot object detection.

2021 — 2021

Developed deep-learning based sensor fusion for driver-assist technologies under the Research & Advanced Engineering group at Ford Greenfield Labs. Implemented a camera-LiDAR fusion model to improve vehicle detection.

2021 — 2021

Contributed to self-driving research for tightly-constrained, mixed-autonomy environments. Visualized data of human driver behavior in parking lots, designed a Python dataset API, and investigated metrics for vehicle interaction.

2020 — 2020

Built out a web app for Trill’s anonymous, supportive social network, developing real-time messaging and friends features. Redesigned onboarding process to increase engagement and inclusion of gender-minority users.

2018 — 2020

Led a group of students on FRC Team 2473 to develop 3D pose estimation algorithms. Applied the Point-n-Perspective problem to detected vision targets to calculate robot position for autonomous navigation.

2019 — 2019

Created an interactive inventory dashboard for Linux systems at the Sherlock Data Warehouse. Automated air traffic data transfers and wrote scripts to process geospatial data.



Contract project for Lightspeed Microscopy developed through Launchpad (launchpad.berkeley.edu). Leveraged generative models and image-to-image translation techniques to perform computational tissue staining.


Used deep and reinforcement learning to train a No-Limit Texas Hold’em poker agent. Experimented with Deep-Q Learning, Neural Fictitious Self-Play, and Counterfactual Regret Minimization.


Explored identification of human-object interactions in images using visual, semantic, and spatial features. Experimented with CNN and GNN architectures as well as incorporating pose data.


Studied tweets to examine xenophobia throughout COVID-19. Collected tweet data and analyzed sentiment to compare correlation with major events.


Created a Chrome extension to make baking more accessible by suggesting ingredient alternatives on recipe websites for those with dietary restrictions.


2022 — 2024

Developing and teaching workshops on machine learning concepts for new members.

2021 — 2022

Helping younger students navigate the EECS community at Berkeley. Providing academic and career advice and support.

2022 — 2022

Taught workshops on web development and Python for a one-week introductory CS program.

2021 — 2021

Worked with a team of high school girls to design and build a mobile app for community upcycling. Taught product and programming concepts, and guided students through creating their first app.



Awarded to top 9% of Teaching Assistants in each department.


Breadth area emphasis in mathematical logic.


Recognition for outstanding academic achievement.