AI-Based Enterprise-Grade CRM
Rohit is excellent to work with and a master of his craft.
I help teams turn difficult AI ideas into reliable products and share what I learn.

Short takes. Real experiments.
AI news, ideas, and things I’m building.
Pressing play loads YouTube and allows its cookies and storage for playback. You can instead use Watch on YouTube. Privacy details.
For projects, use consulting. For anything else, send a note.
Contact and business detailsPreparing contact form…

Check whether llama.cpp b11398’s tinyBLAS K-tail support benefits your CPU vision workload, interpret reported timings and plan a controlled upgrade.

llama.cpp b11393 clears a stale tool-call pointer. See which parser configurations warrant an upgrade and why the earlier b11390 CUDA fix is separate.

llama.cpp b11390 corrects CUDA MoE buffer padding. Identify relevant deployments, pin the fix and validate real workloads without assuming a speedup.