Hi, I’m Marcus, part of the team working on Paddock at Truespar. We’d love to connect with developers in India interested in local AI, Rust and GPU programming.
Paddock runs open models on NVIDIA hardware. It includes a Rust inference engine, our CUDA kernels, OpenAI- and Anthropic-compatible APIs, and a built-in Studio for managing models and comparing local and cloud responses. The code is available under MIT or Apache-2.0, including the kernels.
The latest release, 0.1.6, adds a dedicated DGX Spark build alongside Windows and Linux x64. The project is young, and we’d welcome help making it easier to use and contribute to.
There are several ways to get involved: Rust engine and API work, the Vue/TypeScript Studio, documentation, reproducible bug reports, and GPU testing. The Rust build and regular contributor checks can run without a GPU; kernel work and performance validation need one.
Contributors can request private GPU access for project work: RTX 5090, RTX PRO 6000 or B200, initially for 24 hours, with extensions for regular contributors subject to availability. CONTRIBUTING.md in the repo explains how to request it.
If this is your kind of project, have a look and tell us what you’d like to work on. Feedback on the setup and contribution instructions would also be useful.