The BEHAVIOR Challenge at NeurIPS 2025 invites researchers to tackle long-horizon, everyday household tasks in realistic virtual home environments, supported by a large dataset of 10,000 richly annotated expert trajectories (over 1,200 hours) to advance robot planning and control in complex, human-centric settings.
About Me
I’m a Software Engineer at the Stanford Vision & Learning Lab, where I focus on robotics research. My work centers on developing 3D robot simulations and curating datasets to advance robot learning.
Professional Experience
![]() |
Stanford AI Lab 01/2024 - present Software Developer 2 |
![]() |
Johnson & Johnson 06/2023 - 09/2023 Robotics & Controls |
![]() |
Northwestern Delta Lab 03/2021 - 06/2022 Research Assistant |
Education
![]() |
Northwestern University 09/2022 - 12/2023 M.S. in Robotics |
![]() |
Northwestern University 09/2019 - 06/2022 B.S. with honors in Computer Science, summa cum laude |
Research
MoMaGen automatically generates diverse training datasets for bimanual mobile manipulation by solving constrained optimization problems that ensure robot reachability and camera visibility from minimal human demonstrations.
We introduce the BEHAVIOR Robot Suite (BRS) for household mobile manipulation, featuring a bimanual wheeled robot with 4-DoF torso that achieves critical capabilities in coordination, navigation, and reachability. Our framework includes a cost-effective teleoperation interface and novel algorithm for learning whole-body visuomotor policies.
BEHAVIOR-1K is a comprehensive simulation benchmark for human-centered robotics with 1,000 real-world tasks. Powered by NVIDIA's Omniverse, it features diverse scenes, objects, and activities with realistic rendering and physics simulation. This benchmark aims to advance embodied AI and robot learning research.
Past Projects
Enhancing human–robot collaboration with Omnid Mocobots, this work leverages imitation learning methods – including ACT and Diffusion Policy - to effectively forecast human intent during shared mobile manipulation tasks.
A voice‐controlled robot cooking assistant that integrates LLM for recipe planning, CLIP for object detection, and MediaPipe with LSTM for hand gesture recognition. Users interact with a robot arm through Alexa commands to collaboratively prepare meals.
Developed from the ground up in C++ and simulated using ROS2 and Rviz, this feature‐based EKF-SLAM system employs custom landmark detection algorithms, enabling simultaneous mapping of the environment and precise self‐localization.
By combining computer vision for tower and brick recognition, a custom MoveIt API for motion planning, and fine-tuned MobileNet for hand detection, this solution empowers a Franka Emika Panda robot arm to play Jenga with finesse.
Merging dependency parsing with pre-trained language models through edge-conditioned graph convolutional networks, this approach attains 87.36% accuracy on IMDB binary sentiment analysis using fewer parameters than traditional methods—all while keeping sentence representations both semantic and syntactic.
Here, trajectory planning and control were implemented for a KUKA youBot mobile manipulator—a 5R arm on a four-wheeled mecanum base—to execute an eight-segment pick-and-place routine using feedforward control and odometry-based kinematics simulation for precise object transfer.
Optimizing monocular visual odometry through a region-of-interest (ROI) strategy for feature detection, this solution significantly cuts computational demands while preserving the accuracy usually achieved by processing whole image frames.
Two swarm control algorithms were developed: one mimics the Brazil nut effect to spatially sort robots by size, and the other replicates Reynolds' flocking behavior to coordinate movement reminiscent of bird flocks—all using distributed control techniques.
An autonomous quadrotor drone was built using IMU integration, PID control, joystick input, and Vive lighthouse positioning, resulting in a versatile platform capable of both manual and autonomous flight.
Inspired by Mario Kart, a line-following motorcycle was built using a Raspberry Pi Pico W for image processing and a PIC microcontroller for precise steering control, complete with custom-designed PCBs — earning “Best Design” at the 2023 Northwestern Tech Cup.
A collaborative project where two teams created the IMUGripulator - a 2-DOF robotic arm system combining IMU-based joint control and EMG-based gripper control, built using micro:bit v2 microcontrollers and programmed in C, featuring various sensors including a 9-DOF IMU for tilt-based movement and capacitive touch controls for sensitivity adjustment.
A physics simulation project that models a dice bouncing inside a spinning cup, implementing Lagrangian dynamics to handle the 6 degrees of freedom system, including collision detection between the dice's corners and cup edges, while accounting for gravitational and external forces to maintain the cup's position.
Orchestration Scripts is a framework designed to foster effective work practices. It detects various workplace situations and delivers tailored strategies by abstracting organizational processes, structures, venues, and tools into computational concepts.
Enhancing facial expression detection for individual users, iExpressionNet employs a two-stage approach. Initially, a CNN is trained on the general FER-13 dataset and then fine-tuned using personalized expression data — with frozen convolutional layers — all integrated with OpenCV for robust face detection.