NVlabs/kda
Kernel Design Agents is an agent-centric workflow for using coding agents to research, implement, verify, and iterate on CUDA kernel tasks.
Collecting fresh signals — velocity needs a few days of history.
collecting data…
star history
KDA provides a documented workflow and prompt templates that orchestrate coding agents to handle performance-sensitive CUDA kernel development. It includes repository-facing agent instructions, a generic starter prompt, and a community wishlist where users submit kernel optimization requests for NVIDIA B200 and B300 GPUs. The project is positioned as an early research prototype from NVIDIA Labs.
Frequently asked
- What is NVlabs/kda?
- Kernel Design Agents is an agent-centric workflow for using coding agents to research, implement, verify, and iterate on CUDA kernel tasks.
- Is kda open source?
- Yes — NVlabs/kda is an open-source project tracked on heatdrop.
- How popular is kda?
- NVlabs/kda has 1k stars on GitHub.
- Where can I find kda?
- NVlabs/kda is on GitHub at https://github.com/NVlabs/kda.