I’m a robotics engineer focused on control, optimization, and autonomy — building systems where model-based methods and learning meet.
I've just graduated with a Masters in Electrical Engineering at NYU, working on multi-modal anomaly detection for autonomous vehicles and building high-fidelity simulation pipelines across CARLA, Isaac Sim, and MuJoCo. Meanwhile, also spearheading projects on the kart we've built! 
I care about one thing:
making autonomous systems behave reliably in the real world — not just in clean simulations. That said, here is my website: sous_chef
-
🚗 Autonomous Driving Research (ASAS Lab)
Multi-modal anomaly detection using VLMs + statistical models: (two papers out!) → semantic reasoning (Cosmos) + internal fault detection (Isolation Forest)
→ evaluated in CARLA with ISO 26262-style safety framing -
🧠 Embodied AI for Edge Cases
Defining “external anomalies” as semantic constraint violations
→ bridging perception → reasoning → downstream planning -
🛰️ Space Robotics Simulation
Multi-satellite debris containment (6-DOF dynamics, CW models, cyclic pursuit)
→ validated across analytical + MuJoCo physics -
🐕 Legged Locomotion (Isaac Lab)
PPO-based torque-level control for Unitree Go2
→ actuator-aware training, symmetry-based reward shaping -
🤖 Multi-Robot Systems (ROS 2 + PX4)
Decentralized flocking across aerial + ground robots
→ real-time control with communication + sensing constraints
Core
Robotics & Infra
Simulation & Autonomy
- How learning-based policies + optimal control should coexist (not compete). And I'm exploring this using NMPC based tandem drifting!
- Behavior simulation as the real bottleneck in autonomy stacks
- Pushing simulation → real transfer without cheating physics
- Real-time constraints in VLM-based reasoning systems
I write to understand things deeply — usually at the intersection of:
- control theory
- reinforcement learning
- real-world robotics systems
Recent work:
- VLM-based anomaly detection for AVs (Cosmos-Reasoning)
- Decentralized swarm control under real-world constraints
- Physics-consistent space robotics simulation
If you're working on autonomy, manipulation, or simulation-heavy robotics — I’m always up for a serious conversation.
The best robotics systems aren’t purely learned or purely modeled —
they’re engineered at the boundary between the two.