🤓 Latest Submissions
OpenDine
OpenDine delivers an end to end, closed loop Physical AI system for autonomous bimanual table preparation inside the MuJoCo 3 physics simulation engine. The platform coordinates dual 6-DoF SO-101 robotic manipulators to execute a complete 18 stage physical manipulation workflow conditioned on natural language instructions and multi-camera inputs (an overhead workspace view and dual in-hand wrist streams). Key coordination milestones include center plate positioning, organizer cutlery retrieval with coordinated in-air center hand-offs, stabilizing a beverage bottle to unscrew its cap 90° counter-clockwise, and streaming fluid directly into a ceramic mug before retracting to standby. The underlying SmolVLA policy combines a SigLIP vision encoder, SmolLM2-500M language backbone, and a flow-matching action head. Optimized for edge robotics with Intel OpenVINO, the pipeline utilizes an NNCF INT8-quantized vision tower compressed 3.9× down to 25.8 MB with minimal action drift (0.0310 rad). Benchmarked inside an Intel ready Podman container, the CPU runtime achieves 289.34 FPS with a 3.03 ms mean latency, delivering over 28× headroom above the 10 Hz simulation loop. Evaluated across 10 randomized seeds with mass, position, and surface friction perturbations, OpenDine maintained a 100% task success rate. The project is delivered alongside an interactive browser-based app featuring native 3D drag-to-orbit camera controls, dynamic precision toggling, and live hardware telemetry.
16 Sep 2026
Torch is Lit
"Torch is Lit 🔥❤️🔥" is an innovative project that combines the power of Retrieval Augmented Generation (RAG) and Large Language Models (LLMs) to create a comprehensive solution for developers and researchers. The project consists of two distinct interfaces: "Lit Torch" and "Lit Code." "Lit Torch" is a RAG application that leverages the PyTorch documentation and examples as its corpus data. Built using cutting-edge technologies like Vectara and LlamaIndex, this interface allows users to ask natural language queries related to PyTorch and receive relevant information from the corpus, along with generated responses from the LLM. This powerful combination enables users to gain deeper insights and understanding of PyTorch's functionalities, making it an invaluable tool for researchers and developers working with deep learning frameworks. The second interface, "Lit Code," is a coding-tuned LLM hosted on Together.AI, a platform dedicated to facilitating collaborative coding experiences. This interface is specifically designed to assist users with programming tasks, providing intelligent suggestions, code generation, and debugging support. By leveraging the power of LLMs, "Lit Code" can understand natural language queries and provide relevant code snippets, explanations, and solutions, making it an indispensable tool for developers of all levels. The synergy between the two interfaces in "Torch is Lit 🔥❤️🔥" creates a comprehensive and powerful solution for developers and researchers working with PyTorch and coding in general. Whether users need to understand PyTorch's intricacies, generate code snippets, or debug existing code, this project offers a seamless and intelligent experience, empowering users to achieve their goals more efficiently and effectively.
19 Apr 2024
