ToolMind: Embodied Agent Pipeline
An end-to-end agent pipeline that lets a robotic arm perceive, reason about, and manipulate objects using RGB-D sensing and VLA-style planning.
- Embodied AI
- VLA
- RGB-D
Selected research and engineering projects. Links to code and demos are placeholders and will be filled in as repositories go public.
An end-to-end agent pipeline that lets a robotic arm perceive, reason about, and manipulate objects using RGB-D sensing and VLA-style planning.
An embodied-agent pipeline for a tabletop robotic arm: RGB-D perception builds a scene representation, and a vision-language-action planner turns natural-language instructions into executable motion primitives.

A training-free inference framework that repeatedly refines intermediate hidden states in frozen flow-matching image generators.
See the project page for the method overview, main results, and qualitative comparisons.
Practical deployment of large-model inference services with SGLang: quantization, scheduling, and throughput-oriented serving on limited GPU budgets.
Hands-on serving of large language and multimodal models with SGLang: quantization, continuous batching, and latency/throughput benchmarking on modest GPU hardware.