<feed xmlns="http://www.w3.org/2005/Atom"> <id>https://bhattarai-b.github.io/</id><title>Bibek</title><subtitle>I write about whatever dumb stuff is consuming my mind at a time.</subtitle> <updated>2026-05-01T07:58:00+00:00</updated> <author> <name>your_full_name</name> <uri>https://bhattarai-b.github.io/</uri> </author><link rel="self" type="application/atom+xml" href="https://bhattarai-b.github.io/feed.xml"/><link rel="alternate" type="text/html" hreflang="en" href="https://bhattarai-b.github.io/"/> <generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator> <rights> © 2026 your_full_name </rights> <icon>/assets/img/favicons/favicon.ico</icon> <logo>/assets/img/favicons/favicon-96x96.png</logo> <entry><title>Building a Production-Grade GEMM Library with oneDNN's BRGeMM Micro-Kernels</title><link href="https://bhattarai-b.github.io/posts/brgemm/" rel="alternate" type="text/html" title="Building a Production-Grade GEMM Library with oneDNN&amp;apos;s BRGeMM Micro-Kernels" /><published>2026-04-04T17:42:00+00:00</published> <updated>2026-05-01T07:57:31+00:00</updated> <id>https://bhattarai-b.github.io/posts/brgemm/</id> <content type="text/html" src="https://bhattarai-b.github.io/posts/brgemm/" /> <author> <name>Bibek Bhattarai</name> </author> <category term="Blogging" /> <category term="Tutorial" /> <summary>Part 2 of the “Accelerating Deep Learning on Modern CPUs” series Recap: Where We Left Off In Part 1, we walked through the evolution of matrix multiplication on CPUs — from a naive triple loop (730ms) through loop reordering (130ms) and tiling (89ms), all the way to AMX’s 2D tile registers that can deliver 1024 BF16 ops/cycle. But hand-coding AMX tile intrinsics (_tile_loadd, _tile_dpbssd,...</summary> </entry> <entry><title>A Short Intro of Intel Advanced Matrix Extensions (AMX)</title><link href="https://bhattarai-b.github.io/posts/intel-amx/" rel="alternate" type="text/html" title="A Short Intro of Intel Advanced Matrix Extensions (AMX)" /><published>2026-04-04T17:42:00+00:00</published> <updated>2026-04-04T17:42:00+00:00</updated> <id>https://bhattarai-b.github.io/posts/intel-amx/</id> <content type="text/html" src="https://bhattarai-b.github.io/posts/intel-amx/" /> <author> <name>Bibek Bhattarai</name> </author> <category term="Blogging" /> <category term="Tutorial" /> <summary>Based on my talk at QCon San Francisco 2024 Watch the talk: Maximizing Deep Learning Performance on CPUs using Modern Architectures — includes slides, video, and full transcript. Introduction When people think about deep learning acceleration, GPUs are what typically come to mind. But in my experience helping teams deploy AI workloads across cloud, on-premises, and hybrid environments,...</summary> </entry> <entry><title>Designing Retrieval Augmented Generation (RAG) Systems</title><link href="https://bhattarai-b.github.io/posts/rag-systems/" rel="alternate" type="text/html" title="Designing Retrieval Augmented Generation (RAG) Systems" /><published>2025-12-31T17:42:00+00:00</published> <updated>2025-12-31T17:42:00+00:00</updated> <id>https://bhattarai-b.github.io/posts/rag-systems/</id> <content type="text/html" src="https://bhattarai-b.github.io/posts/rag-systems/" /> <author> <name>Bibek Bhattarai</name> </author> <category term="Blogging" /> <category term="Tutorial" /> <summary>Retrieval Augmented Generation (RAG) is one of the most popular techniques to enhance the performance of Large Language Models (LLMs) models by injecting external data in the context. Like the name suggests, it involves retrieving the data from external source outside of LLMs pre-training data, augmenting the prompt with retrieved information, and using this information-rich prompt to enhance t...</summary> </entry> <entry><title>Agent Layer- Building Agentic Workflows to Solve Real Problems</title><link href="https://bhattarai-b.github.io/posts/agentic-workflow/" rel="alternate" type="text/html" title="Agent Layer- Building Agentic Workflows to Solve Real Problems" /><published>2025-12-31T17:42:00+00:00</published> <updated>2026-04-06T05:43:09+00:00</updated> <id>https://bhattarai-b.github.io/posts/agentic-workflow/</id> <content type="text/html" src="https://bhattarai-b.github.io/posts/agentic-workflow/" /> <author> <name>Bibek Bhattarai</name> </author> <category term="Blogging" /> <category term="Tutorial" /> <summary>Disclaimer: This is not a tutorial, book, or a whitepaper. The content on this file are notes I scrambled while learning the Agentic systems. The major resourses I have relied on are Andrew Ng’s course on Agentic AI, as well as the book “Agentic Design Patterns” by Antonio Gulli. In addition to that I have looked into several blog posts, youtube videos, and whole lot of brainstorming with Gemin...</summary> </entry> </feed>
