Hello, I'm ighoshsubho
I'm an MLE, X yapper, and an aspiring AGI researcher.
My work primarily involves around Inference optims, ML research, contributing through OSS tools for diffusion models, and inference engines.
- Blackwell-Kernels - Educational matmul kernels for sm_120a arch gpus.
- Register-Spill-Debug - Vibecoded Register spill debugger in HTML for PTX and SASS.
- Hunyuan-Image2.1-for-GPU-Poor - FP8 optimized hunyuan image on 24 gigs.
- vLLM-chatterbox-Multilingual - vLLM support for multilingual chatterbox.
- MCTS-Distributed - Parallel MCTS with MPI and groq for LLMs reasoning
- Diffusion-Kernels - Kernels for attention and other diffusion-specific tasks.
- Blog: https://ighoshsubho.bearblog.dev/
- Twitter: https://X.com/subhoghosh02
- LinkedIn: https://linkedin.com/in/thesubhoghosh
- Instagram: https://instagram.com/subhoghosh02



