.../Guides...
Field notes on building real software
Direct, practical answers to the questions that come up when you build full-stack apps, RAG systems, and AI-powered products.
RAG vs Fine-Tuning vs Long Context
Use RAG for knowledge that changes, fine-tuning for behaviour and style, and long context for prototypes or whole-document tasks. Here is how to decide.
Read guideHow to Add an AI Chatbot to a Web App
Add an AI chatbot as a bolt-on: put the LLM behind its own service, ground it in your data with RAG if needed, stream responses, and add fallbacks.
Read guideNext.js + FastAPI Architecture Guide
Run Next.js and FastAPI as two services talking over a JSON REST API. Use Next.js for the UI and light edge logic, FastAPI for heavy Python and AI work.
Read guideHow these guides are written
Every guide follows the same answer-first format: the direct answer in the opening paragraph, a TL;DR you can act on, question-style headings, and a short FAQ. The goal is that you leave with a decision made, not a longer reading list.
The advice comes from first-hand production work rather than surveys of other posts — a RAG chatbot built with the OpenAI API and the Qdrant vector database, frontend and integration work on EigenTalk (a document-Q&A research assistant), and Next.js and FastAPI services shipped through GitHub Actions CI/CD. Where a guide recommends an approach, it is because that approach was chosen — or rejected — in a real build.
Three topics are covered so far. The RAG guide weighs retrieval-augmented generation against fine-tuning and long context windows, and explains when each one actually fits. The AI-products guide walks through adding a chatbot to an existing web app without destabilising the rest of the product. The full-stack guide covers structuring a Next.js frontend with a FastAPI backend, from API contract to deployment. If you would rather have the work done than read about it, the services page describes what I build, and the projects behind these guides are documented as case studies.