Aleksandr S.

I build and ship production backends, and the full stack around them when a project needs it. Twenty years in backend and seven in machine learning, the last several building web apps and AI features that real businesses actually use. A few things I've built: Neptyne. A programmable spreadsheet that uses Python as its formula language, plus a Google Sheets add-on that runs Python inside Sheets. I worked on spreadsheet functions, optimizations, and a code debugger. Lekalo.app. A tool that fits complex-shaped parts onto rectangular sheets to cut material waste. A lot of geometry and optimization math on a Python backend, with the heavy work in background jobs. Gigabite (gigabite.ai). Lead engineer for 18 months. Django and django-ninja API, Celery workers, a Next.js front end, all on GCP (Cloud Run, Cloud SQL, Firebase). It does landed-cost and tariff pricing, invoice matching backed by the Claude API, and bulk-data admin tools. It served real foodservice distributors. Day to day I work in Python (Django, FastAPI, Flask) and TypeScript/React, with PostgreSQL, Celery, Docker, and whatever cloud the project runs on. I also work in Rust and C++ when a project needs the performance. On the AI side I build LLM features end to end: RAG, agent and multi-agent systems, and the infrastructure around them (queues, workers, vector search), keeping model calls behind one clean boundary so the provider can be swapped without touching the rest. The ML background runs deeper than LLMs, from model training to computer vision and NLP. I like problems with real constraints, and I'll tell you straight what will work and what won't.