I'm open to work :)
🧑🏻🔬 Senior Machine Learning Engineer 8yrs, Computer Vision, Data Science.
🪪 Solid software engineering skills, Ph.D. in electronic engineering, 3 years as a research fellow before that, co-founded a startup back then.
🔮 LLM, photos, satellite images, IoT, geospatial, weather.
Recent Posts
Of 22 spec formats, agents need only a few: code as the final spec, RFC → implementation plan for features, tests and golden files for verification, and issue-tracker items for everything else.
Should a company conduct war games simulating the shutdown of its data center and all available AI assistants?
Coding is solved. But engineering practices aren’t. What is Claude Code missing? Even for small pet projects, I crafted several documents describing my approach and scripts to enforce it.
Router to use different local LLM models and all Anthropic models simultaneously in one Claude Code session. Saving telemetry for observability and a corpus of all raw requests and responses for later analysis.
One prompt for the screensaver itself and the rest work to create an uninstaller :)
Do you want to keep your business safe? Do you want to protect your data and your customers? Go for local models. Even for coding agents.
First, you want to be the top one guy in burning tokens (Meta-style KPIs). Second, you want to ‘copy’ a well-built project: C++ compiler, PyTorch, etc.
Yes. I believe every IT engineer will eventually have an AI box with 1024 GB of unified RAM, capable of running an average optimized model locally—enough for everyday work.