I am a robotics engineer at Northeastern University. I build systems that can sense, decide, and move, from MPC humanoid locomotion to VLA policies running on real robot arms.
Northeastern University News interviewed me about my work on Space Station OS, an open source ROS 2 platform that simulates life support, thermal control, and electrical power for a station like the ISS.
Read the article →These range from imitation learned VLA policies to MPC based autonomous racing and open source space systems. My goal with each of them was to make it work in practice, on real dynamics or real hardware.
A three layer autonomy stack for a bipedal Unitree G1 on rough Martian terrain. It uses PPO locomotion under Mars gravity, an error state EKF with GTSAM as a "Mars GPS" for localization without satellites, and A* with MPPI for navigation. Everything runs on real NASA Jezero Crater terrain.
View project →A convex MPC locomotion controller for the Unitree G1 humanoid in MuJoCo. It uses single rigid body dynamics, Pinocchio for whole body control, and the OSQP solver at 1.4 ms. I fixed 12 integration bugs before it walked steadily, then grew it into HANUMAN for rough Mars terrain.
View project →An end to end multimodal policy I built from scratch. It has two camera ViT encoders, a PaliGemma tokenizer, a 4 layer Transformer, and 50 step action chunking. On a real 7 DoF arm it picked up a screwdriver with a 73% success rate.
View project →MPC trajectory tracking and MPCC racing line control with CasADi and IPOPT, with dynamic rollout planning using RK4. I ran it on a real Ackermann car.
View project →
RL trained locomotion in Isaac Lab combined with ROS 2 Nav2 for indoor navigation without GPS, set in disaster style environments.
View project →
A modular ROS 2 platform that simulates life support, thermal control, and electrical power for a station like the ISS. Featured by Northeastern University.
View project →A study that compares missile autopilots across SFI, LQR, and MPC on short period longitudinal dynamics. MPC reached the target fastest at 8 s and used the least control effort.
View project →
A classical pipeline with HSV perception, Pinocchio IK, and OMPL RRTConnect. Later I moved to a PPO policy trained in MuJoCo for more adaptive pick and place.
View project →
An outdoor delivery robot for last mile logistics in neighborhoods, with autonomous navigation, obstacle avoidance, and delivery coordination.
View project →I am a robotics engineer at Northeastern University and I work across locomotion, manipulation, autonomy, and space systems. What I care about most is building control systems that hold up under real dynamics, not only in a clean demo.
So far I have built MPC locomotion controllers for humanoid robots, run VLA policies on 7 DoF arms, raced autonomous cars with MPC and MPCC, contributed to open source space station software, and trained legged robots for disaster response.
I am open to research collaborations, engineering roles, and any chat about autonomy, manipulation, or space systems. Feel free to book a coffee chat or just send me a message.
[email protected]