About Me
AI Engineering and Inference Optimization Specialist
Yet Another AI Blog
Last 10 notes on Yet Another AI Blog
Latest posts
Interactive Tools
Hands-on tools for understanding GPU programming and AI optimization
Hermes Research with khoj-cli
How a small Exa-powered terminal search tool becomes much more useful when wrapped as a Hermes skill.
Qwen3.5 on Dual R9700s
Benchmark notes from dual-R9700 ROCm inference: good warm throughput, decent TTFT, and a tensor-parallel path that still likes to explode under stress.
LLM-Guided Kernel Optimization
A technical discussion of how large language models can speed up GPU kernel optimization, from research ideas to production kernels.
Triton Linear Layouts
Exploring the concept of linear layouts in Triton, a programming language for writing high-performance GPU code, and how it simplifies memory management and data access patterns.
PyTorch 2.x Backends
PyTorch 2.x Backends The Great ML Framework Debate Centuries ago (2020), I contributed to PyTorch library, specifically, TorchServe which used to be the default model serving library for PyTorch Models.
CUDA Threads and Blocks
An in-depth exploration of CUDA configurations and their impact on thread/block patterns for optimal performance
CUDA Thread/Block Visualizer
CUDA Thread/Block Visualizer This interactive tool helps you understand how different thread and block configurations affect GPU execution patterns.
CUDA Thread/Block Visualizer
CUDA Thread/Block Visualizer This interactive tool helps you understand how different thread and block configurations affect GPU execution patterns.