README.md
reterraform.antresspace_dust.antresthe_junkyard_of_the_jettisoned.antresapte.md42-school.mdexperience.mdskills.md
hello-world.htmlMy daily language for 6+ years. I write everything typed and run mypy in strict mode. I like Python when it's treated as a serious language — type hints, dataclasses, protocols. I don't do magic imports or untyped dicts passed around.
Python mypy FastAPI Pydantic SQLAlchemy Celery Alembic
I aim for high coverage not as a metric but because I don't trust code that isn't tested. That's also why I'm building ProTest — I wanted DI in tests to be explicit and typed, not resolved by name in the dark.
pytest ProTest
PostgreSQL for relational work, comfortable with raw SQL when ORMs get in the way. But what really excites me is graph databases — not just as a storage alternative, but for what you can do with them: shortest path, community detection, and the algorithms that come with thinking in graphs. Currently exploring neuro-symbolic approaches and world models on a personal project (Felix), where graph structure meets scenario extraction.
PostgreSQL Neo4j Cypher Kuzu ChromaDB Weaviate
Docker for everything local, Kubernetes and AWS in production at work. My personal projects stay simple — a static Nuxt front and Supabase is usually enough.
Docker Kubernetes AWS RabbitMQ GitHub Actions
I orchestrate, build agents, and pick the right model for the job — RAG pipelines, structured extraction, knowledge bases. What I'm most excited about right now is making AI systems that run on a normal computer: small, focused models with targeted tools and a lot of heuristics behind them, rather than throwing everything at a giant model. It reminds me of writing C — you can't brute force it, you have to be clever.
vLLM Pydantic AI LM Studio
I don't love it. I don't love JS, I don't get why anyone would run it on a backend, and the layers of retro-compatibility annoy me. And yet — web technologies work remarkably well. The ecosystem is a genuine success on many levels, and I respect that. This is where I vibe-code: I want a result, not a journey. The opposite of how I approach Python or C. I still try to follow good practices, even if sometimes I end up writing text scrambling effects under the user's cursor.
TypeScript Vue Nuxt
C and C++ from 42 — I don't write them daily but they shaped how I think about memory and performance. Comfortable in Linux. PyCharm for everything Python.
C C++ Linux PyCharm
Want to get better at
Graph databases — I use Neo4j and Kuzu but I want to write Cypher more fluently and go deeper. Rust — I've wanted to learn it for a while but never had the right project for it. Networking and systems — I know the basics but it's not where I'm efficient yet. The internals of AI — I orchestrate models daily but I want to understand what happens under the hood, neural networks for real. And English — I can write it (with some help) but I can't hold a conversation comfortably. Working on that by end of 2026.