Jobanpreet Singh Nagra

RoboGPT is a robotics-focused AI assistant built using a Retrieval-Augmented Generation (RAG) architecture.
It leverages high-quality vector embeddings to enable low-latency, semantic retrieval of robotics knowledge.

Core Technology: Embeddings + RAG

The system uses sentence-transformer embeddings with ChromaDB as the local vector store. This eliminates network latency and speeds up semantic retrieval during query resolution.

Context Window Optimization

Instead of fine-tuning a robotics-specific LLM, domain knowledge (ROS2 docs, robotics textbooks, Python robotics resources) is injected into the model context on demand. This provides high accuracy without additional training cost.

Hybrid Compute Strategy

Lightweight embedding computation is performed locally, while heavy reasoning is delegated to the DeepSeek R1 API.
This balances throughput, latency, and cost while delivering high-precision robotic task guidance.