Deploying Lumina: A Scalable AI Photo App on Hetzner & Coolify
How I deployed a complex monorepo stack for $4/month using Coolify, while overcoming OOM crashes and configuration drift.
Software Engineer learning to specialize in distributed systems and performance optimization. Currently building at Flutterwave.
I write on some of the most interesting things I am learning and/or unlearning. New experiences, old ones, and the 🎧 music I like.
An educational project implementing internal storage engines from scratch. Features an LSM-Tree based Key-Value store and a JSON Document store with secondary indexing and MongoDB-style query operators.
A local-first movie recommendation engine and watchlist manager. It builds a personalized taste profile by analyzing Letterboxd history and uses semantic search to find similar films from the TMDb database.
A secure log management and background job processing engine. Features real-time log ingestion, advanced filtering, and encrypted storage with a dynamic HTMX-powered dashboard.
Lumina is an AI Intelligence Layer for your photo library. It allows users to organize photos, perform advanced face recognition, and host Collaborative Events. The platform separates AI compute from physical storage, offering a Bring Your Own Storage (BYOS) model that gives users full control over their data while leveraging powerful facial recognition and search.
How I deployed a complex monorepo stack for $4/month using Coolify, while overcoming OOM crashes and configuration drift.
Designing a distributed face-matching system using Bun, FastAPI, and PostgreSQL—without relying on pgvector.
A technical post-mortem on building a multi-layered relational engine, from byte-level storage to SQL interfaces and ACID transactions.