Building a Cost-Ordered Image Reading Cascade for PDF RAG
A cost-ordered cascade — cheap filter, type check, OCR, vision model — turns only the PDF images worth reading into searchable text for RAG.
Machine Learning
Loss Functions Explained: How Machine Learning Models Know They're Wrong
Loss functions give ML models a measurable mistake score during training. This beginner-friendly guide covers MSE, MAE, cross-entropy, and the training loop.
Parsing Scanned PDFs for RAG with EasyOCR: Text Without Layout
EasyOCR recovers text from scanned PDFs but provides no layout structure. Here's what that gap means for enterprise RAG pipelines.
Ten ChatGPT Features and Techniques That Improve Response Quality
Most users interact with ChatGPT like a search engine and miss its full capabilities. These ten techniques close that gap.
Logits, Temperature, and Top-P in LLM Token Selection
Learn how logits, temperature, and top-p sampling work together to control next-token prediction in large language models.
Building a Multi-Tool Gemma 4 Agent with Error Recovery
Learn how to build a resilient Gemma 4 agent loop that handles tool failures, malformed outputs, and unavailable services using structured error recovery.