A deep research agent built with LangGraph, Qwen3-30B running locally via LM Studio, and full execution observability through Langfuse. This post walks through four patterns that matter for production agentic systems: multi-agent orchestration, the human interrupt pattern, clean context management, and agent observability—all running at zero API cost.
Healthcare and longevity have always fascinated me. This project combines my passion for evidence-based health with AI—building an intelligent system that answers health questions using actual PubMed research papers, powered by Neo4j knowledge graphs and local LLM inference.
In this project, I automated the extraction of ingredient details such as name, unit, quantity, and price from invoices using Azure's Form Recognizer and GPT-4o-mini. The solution leverages state-of-the-art AI models to improve accuracy when data is incomplete or not recognized by default. In this article, I'll explain the choices behind the tools I used, the workflow, and the cost breakdown, making it easy for others to replicate and scale a similar solution.
Learn how I built a professional blog using Next.js and Azure Static Web Apps for just $10/year. This guide covers the file-based architecture, static site generation benefits, and step-by-step implementation that delivers enterprise-grade features at unbeatable cost.