Frontend
Next.js 14, TypeScript strict, Tailwind. WebSocket bridges, in-browser depth-map decoding, multi-camera tabs, live skill registry CRUD. Deployed to Cloudflare Pages.
RobotApp is a personal portfolio piece that also happens to be a real, working robotics product. It is the UI layer of a four-package stack that turns any ROS2 mobile manipulator into a network-addressable agent — driveable from a browser, from a CLI, or from a Python script, all sharing one runtime.
The reference robot (kcare_robot) is an
assistive 6-DOF cobot with two RGB-D cameras, a Nav2-driven mobile base, and
an open-vocabulary perception pipeline (GroundingDINO + GroundedSAM +
mask2grasps). It’s deployed in real environments today.
Frontend
Next.js 14, TypeScript strict, Tailwind. WebSocket bridges, in-browser depth-map decoding, multi-camera tabs, live skill registry CRUD. Deployed to Cloudflare Pages.
Backend runtime
FastAPI service (~4.2 K LOC) with custom JSON encoder for numpy / PIL, atomic-persistence layer, async-to-thread WebSocket bridge, cross-mode bootstrap.
Robotics integration
ROS2 Humble integration via pyconnect; lazy node init with 4 callback
groups, Nav2, action / service / topic clients, head-to-base
calibration, parallel actuator coordination.
Vision
Open-vocabulary detection pipeline over TCP, 3D feature extraction (cluster normals, inverse projection, lying-object classification), grasp-pose lifting (2D pixels → 6-DOF tool pose).
Cookiecutter scaffold
robot_template — generates a dashboard-ready robot package in seconds,
with three reference patterns for skill implementation.
DevOps
Cloudflare Pages + custom domain (robot.aistations.org), GitHub Pages docs (this site, Astro Starlight), Makefile-driven deploys.
| Repo | What it is |
|---|---|
| robotapp | This dashboard — Next.js + Cloudflare Pages |
| robot_agent | FastAPI runtime — three-mode core |
| kcare_robot | Reference robot — 23 production skills |
| robot_template | Cookiecutter scaffold for new robots |
Trung Bui · Robotics + Vision Engineer
Open to roles in robotics, computer vision, and full-stack AI infrastructure. Happy to walk through any layer of this stack on a call.