News
From Local LLM to Tool-Using Agent
4+ hour, 46+ min ago (905+ words) Using Gemma 4, Ollama, Open AI Agents SDK, and Tavily MCP to build a lightweight research agent You have just deployed a local LLM. Nice. But after the first few chats, you might be wondering: what else can I do with it?...
Water Cooler Small Talk, Ep. 11: Overfitting in RAG evaluation
6+ hour, 16+ min ago (761+ words) Why memorizing for the exam doesn't mean you understand the subject So, here's the water cooler opinion of today's post: We've built a RAG app that is playing out really well. We are now in the evaluation stage, and it's…...
Amplify the Expert: A Philosophy for Building Enterprise RAG
7+ hour, 46+ min ago (1007+ words) Enterprise Document Intelligence [Vol. 1 #M1] " The thesis behind every architectural choice in this series This article is a manifesto of Enterprise Document Intelligence, a series that builds an enterprise RAG system from four bricks: parsing, question parsing, retrieval, and generation. Amplify…...
Vector RAG Isn't Enough " I Built a Context Graph Layer for Multi-Agent Memory
1+ day, 2+ hour ago (1781+ words) A structured context graph beat a flat chat dump and a vector-only RAG pipeline on the same multi-agent conversations " full working code, real benchmark numbers, zero API calls I built a three-agent pipeline that worked great for short tasks. But…...
The Hot Path Belongs to GBDTs, Agents Own the Cold Path: A Payment-Fraud Benchmark
1+ day, 3+ hour ago (1349+ words) A reproducible benchmark on latency, cost, and reproducibility, and where agents actually earn their keep. This article tests a question that keeps coming up in payments work: can an LLM agent replace a gradient-boosted scorer on the synchronous payment authorization…...
Letting an LLM Pick the Right RAG Page: The Arbiter Pattern at the End of Retrieval
1+ day, 7+ hour ago (1742+ words) Enterprise Document Intelligence [Vol. 1 #7 C] " One LLM call ranks the candidates with reasons. The output is one typed object your auditor can defend Retrieval is filtering on line_df and toc_df, anchor versus context: the mental model from Article 7 A (retrieval as filtering)....
Beyond the Straight Line: Choosing Between OLS, Interaction Terms, and Tweedie Regression
1+ day, 4+ hour ago (1475+ words) Mastering GLMs: When to Use OLS, Interaction Terms, or Tweedie Regression for Skewed Data Imagine you are building a machine learning model to predict how much money customers will spend on an e-commerce platform over the next year. The job…...
One Month Into Learning Data Engineering in Public: Here's What I Didn't Write About
1+ day, 9+ hour ago (901+ words) A reflection on the first month of learning data engineering in public, and what actually kept me going. I read that message twice. Because two weeks before that, I had nearly talked myself out of the very roadmap I'd publicly…...
Python 3. 14 and its New JIT'Compiler
1+ week, 3+ hour ago (852+ words) A technical overview and some benchmarks The release of Python 3. 14 marks an important point in the evolution of the world's most popular programming language. While Python has long been acknowledged for its readability and large ecosystem, its execution speed has…...
Building a Custom GStreamer Plugin for NVIDIA Deep Stream
1+ week, 4+ hour ago (1127+ words) Why custom inference in Deep Stream? However the common case has limits. Vision-language models, custom post-processing, rotated bounding boxes, or the need to hot-swap models at runtime, these are places where nvinfer's assumptions break down. Sometimes you have a mature…...