06 / Interests

What I'm thinking about.

Where robotics is heading next, and the questions I keep coming back to.

Current focus

Agentic robotics

Robots as agents, not just policies. A reasoning model plans, calls skills and tools (perception, navigation, manipulation primitives, ROS services), checks its own progress and recovers when things go wrong. Where do agent loops make robots more capable, and where do they just add latency?

  1. goal“put the red cup in the sink”
  2. planfind cup → grasp → place in sink
  3. calldetect(“red cup”)✓
  4. callgrasp(cup_1)✗ slipped
  5. verifyre-plan grasp from new view✓
  6. callplace(cup_1, sink)✓
Current focus

System 0 / 1 / 2

A layered control stack built on fast and slow thinking. The interesting questions are at the seams: what each layer learns, how often it runs and what data it needs.

  1. S2Reason & planVLM / LLM~1 Hz
  2. S1See & actvisuomotor policy (VLA)~10–100 Hz
  3. S0React & stabilizelow-level / whole-body control~1 kHz

Data scaling for robot learning

Which data actually moves the metrics, how performance scales with it, and where the next labeled hour should go.

Embodied foundation models

VLAs, world models, action chunking and flow-matching policies, and the data and infrastructure they really need.

Learning from humans

Egocentric and third-person human video as robot data, and how to bridge the embodiment gap.

Toolbox

What I build with.

robot software

ROS / ROS 2 · C++ and Python nodes · lifecycle & QoS · MCAP / rosbag · bringup · Docker · NVIDIA Jetson

sensing & perception

Camera & LiDAR calibration · SLAM / VIO · point clouds (PCL) · RGB-D & stereo · IMU fusion · YOLO · U-Net · SAM

robot learning

VLA & imitation learning · LeRobot · RLDS / Open-X · VLMs · CLIP / SigLIP · PyTorch · ONNX / TensorRT · quantization

platform & infra

Kubernetes · Airflow · Ray · FastAPI · PostgreSQL · DuckDB · BigQuery · GCP · AWS · dataset versioning & lineage

languages

Python · C++ · TypeScript / JavaScript · SQL · MATLAB