NeuralCube — Autonomous Robot
A personal research robot orchestrated by multi-agent LLMs and built on ROS2. As the backend-focused contributor, I deployed vLLM and SGLang inference servers on Linux for low-latency LLM responses in real-time decision-making. I integrated the multi-agent orchestration layer with ROS2 nodes, routing tasks (navigation, memory, search) to specialized subagents via ROS2 topics and services. I also managed the robot's digital twin state using ROS2 parameters and state publishers, and established documentation standards and systematic debugging practices across our Git workflow.