PinnedBrajendra Singh·Oct 14, 2024Choosing the Right Chunking Method for Effective RAG in Amazon BedrockRetrieval Augmented Generation (RAG) is an architectural approach used in building generative AI applications to improve the accuracy of…
Brajendra Singh·2d agoLoRA Fine-Tuning Made Simple: Concepts and Math ExplainedUnderstand how LoRA works, why it needs fewer trainable parameters, and how lightweight adapters let one LLM handle multiple tasks.
Brajendra Singh·Oct 26AI Models That Decide, Not Write.For years, almost every AI headline has focused on models that write — drafting our emails, summarizing documents, and generating code…
Brajendra Singh·Oct 19AG-UI: The USB-C for AI Agents and FrontendsAgents are easy to build and hard to connect. The backend logic works, but wiring it to a live UI means handling streaming text, tool…
Brajendra Singh·Sep 28What Six Versions of Claude Reveal That No Single Release DoesIn about thirty months — from Claude 3.0 in March 2024 to Claude 5.5 in September 2026 — Anthropic Claude went from “a smart chat-bot that…
Brajendra Singh·Sep 13Skip the Training: Zero-Shot Tabular Classification with MitraIn two of my previous posts, I’ve been exploring a simple but powerful idea: what if you didn’t have to train a model from scratch?
Brajendra Singh·Jul 24What If You Didn’t Have to Train a Regression Model from Scratch?For more than a decade, gradient-boosted trees — especially XGBoost, LightGBM, and CatBoost — have been the gold standard for tabular…
Brajendra Singh·Jul 10OpenAI Agents SDK × Amazon Bedrock Mantle: A Notebook-First Quick StartEver wished you could use the elegant developer experience of the OpenAI Agents SDK while keeping model invocation inside Amazon Bedrock…
Brajendra Singh·Jul 10Deterministic Evaluation for AI Agents: Trusting What Happened, Not Just What Was SaidAgent evaluation is the process of determining whether an AI agent is actually doing a good job. An evaluator analyzes the agent’s behavior…A response icon1A response icon1
Brajendra Singh·Jun 28Agents with Expertise: When Skills Meet Code InterpreterWhat if you could take a domain expert’s years of knowledge — their methodologies, decision criteria, edge-case handling — and bottle it…
