about me 9 blog resume

Gloria Zhu

Hi! I'm a Masters student at MIT ('27) studying robot learning with a focus on manipulation and RL. I want to work on robots that make the world a better place - building homes, providing elder care, assisting in surgeries, etc.!

I finished my undergrad at the 'tute this past spring, studying electrical engineering and computer science (6-2) and design (4-b).

Previously, I worked on trajectory planning at Zoox (Amazon Self-Driving), networking infrastructure at Databricks, and embedded systems at Formlabs.

Research

Improbable AI Lab
(June 2026 - Present)

Currently, I'm investigating real-world RL techniques for dexterous manipulation, particularly for contact-rich tasks.

So far, I've:
e trained and deployed force-aware manipulation policies using different flavors of imitation learning (ACT, diffusion, HG-DAgger) on real hardware
b brought up a software/hardware testing bench for robust demo collection, policy rollouts (and interventions), and sensor fusion on a Flexiv Rizon 4s arm

Biomechatronics Lab
(2023-2024)

I contributed electronics design, signal processing/analysis, writing, and diagrams to a tissue/bone-integrated powered prosthetic knee project, and aided with human subject testing and data collection.

Shu, T., Levine, D., Yeon, S. H., ... Zhu, G., ... Herr, H. M. (2025). Tissue-integrated bionic knee restores versatile legged movement after amputation. Science, 389(6756), eadv3223.

Projects & Blogs

Real-World RL is a UX Design Problem (Lessons from HIL-SERL)

Monte Carlo localization and path planning in the CS building basement

Making a 5-DOF calligraphy robot in one month (CAD -> Electronics -> Code)

Other projects:
- a distributed key/value service built in Go, on top of a full implementation of the Raft consensus algorithm
- a compiler for Decaf (simplified C-like language) -> x86-64 assembly, with dataflow optimizations and register allocation

Internships

Zoox
(Amazon Self-Driving)

Team: Vehicle Behavior (Costing), Summer 2024

I designed, implemented, and analyzed reward functions for autonomous vehicle behavior around school buses, ensuring correct behavior in 30 resimulated real-world traffic scenarios.

Databricks

Team: Networking Infrastructure, Summer 2025

I built and deployed a network probing service in Go to monitor success rates of ingress network requests, providing coverage to previously unmonitored cloud-to-cloud pathways.

Formlabs

Team: Embedded Software, Fall 2024

I designed, prototyped, and built a feature called “Motor Moves” end-to-end in C++ and shipped it in the 2024 Fall firmware release to all flagship Form4/4L 3D printer units.