I push tools past what they were built for.

That's where the real limits are. Usually the simplest thing still standing is the answer.

Lately that's detections in security: detecting phishing emails at Cisco, detecting Kremlin talking points in Czech comments online. Before that, a browser CLI for AI agents, and whatever's on the soldering bench.

Prague · building, and open to collaboration

github linkedin X

builds

kremlometr.cz pro-Russian propaganda in Czech news comments. live since Jan 2026. more builds the full archive: research, experiments, bots →

tools

brow a CLI that lets AI agents drive a real browser. browser automation, pushed onto agents. more tools interactive web tools and CLIs →

notes

30 May 2026 Two Independent Post-Quantum PGP Implementations Now Interoperate 26 Jan 2026 The "Zoo" of Performance Metrics: A Complete Guide to the Confusion Matrix 23 Jan 2026 Something on productivity and tasks in my life 20 Jan 2026 My journey to Machine Learning Engineering 16 Jan 2026 My take on LLM coding 13 Nov 2025 Something little on group testing all notes →

work

Cisco Senior ML engineer on anti-phishing - the engine deciding if an email is trying to hurt you, at enterprise scale.

about

I'm Daniel, in Prague. I studied theoretical biophysics: Redfield equations, open quantum systems and the abstract thinking is the sharpest tool I kept.

Away from the keyboard I climb: bouldering, outdoor sport climbing, occasionally multipitch climbing in the Alps.