Blog
What I learn while building
Notes from working on AI, software and servers. Usually written after finishing something, to remember what actually helped.
Posts8
July 23, 2026
An AI that answers on WhatsApp: from conversation to a real order
Not a bot with canned answers but a tool-using agent: a WhatsApp/Instagram assistant on LangGraph + FastAPI that creates real orders while it talks, runs channel-agnostically and hands off to a human when it should.
Read more →
June 10, 2026
From One Server to Multi-Service: Splitting with Docker Compose
How I split this site from a single app into three services — database, API and frontend — and bring them all up with one command.
Read more →
June 10, 2026
The BFF Pattern: Wiring Next.js to FastAPI Securely
The browser never sees the backend. I explain the Backend-for-Frontend pattern: keeping the JWT in an httpOnly cookie and talking to the API only from the server.
Read more →
June 10, 2026
Hardening My Personal Site: Sanitization, Cookies and JWT
The old version had a vulnerability. While closing it I rebuilt the site securely: HTML sanitization, bcrypt, httpOnly cookies and security headers.
Read more →
June 10, 2026
From camera to alert: real-time detection with YOLO + Supervision
How I built the detection → tracking → rule-violation → alert chain from a camera stream in SafeEye.
Read more →
June 10, 2026
An AI that texts like a human: memory, mood, and "typing…"
What makes an LLM feel "alive" isn't its intelligence; it's persistent memory, mood, and a human-like behavior layer. From BUDDAI.
Read more →
June 10, 2026
When one LLM isn't enough: multi-agent systems with LangGraph
Cramming a task into one giant prompt is brittle. Orchestrating specialized agents with a LangGraph state graph.
Read more →
June 10, 2026
From Compose to CI/CD: automated deployment with GitHub Actions
Wiring a multi-service app (this site: Next.js + FastAPI) into a build → test → image → deploy pipeline on every push.
Read more →