PinnedRaghu Narayan Koratagere·Aug 7LLM-Infra-Cost Simulator — You dont Buy Eight H100s to Find Out If You Need Eight H100sA planner and a simulator for the question nobody answers before the invoice arrives
Raghu Narayan Koratagere·Aug 7How a router is clear option for modelsWe have tried various options w.r.t AI models for inference. Frontier APIs, OnPremise DGXs, Cloud GPU providers and finally optimised…
Raghu Narayan Koratagere·Jul 22Coding Model setup for Engg TeamI was looking at setting up locally hosted models for coding for our Engg Team of 150 engineers.
Raghu Narayan Koratagere·Jul 21Gemma E4B on Mac Pro (16GB)Out of curiosity, I profiled an open-model, gemma-4-e4b-it a multimodal (text, video and audio ) model on my local Mac Pro 16GB — since i…
Raghu Narayan Koratagere·Jul 8DGX Spark — What can i do with itIts been a season for running LLMs locally — top reasons were Cost Savings, Data/Code security, Control, Nearness to Repo etc.
Raghu Narayan Koratagere·Jun 11What a Cache ! — the Gemini catchupWe figured that Gemini models use implicit caching based on the stable byte prefix of requests. What this means, is model does best effort…
Raghu Narayan Koratagere·May 23What a Cache !We Blew Through Our AI Budget in Two Weeks. Here’s What We Learned.
Raghu Narayan Koratagere·May 21Antigravity 2.0 — Is it a Big Deal !Goggle I/O’s announcement of AnitGravity 2.0 has caused so much excitement in developer community.