Writing

Notes & experiments

Working notes on speech models, training dynamics, reinforcement learning, and the occasional side quest.

Qwen3-TTS for Swiss German: Dialect Adaptation via Talker Layer Fine-Tuning

My Personal AI: Living with OpenClaw

Numerai – Top 10 Participant and Numerai Master

The Case of the Diverging Loss: When 'Best Loss' != 'Best Model'

Visualizing Whisper: Mapping Cross-Attention to Time

Solving the game "Overcooked" with Reinforcement Learning

SwissText 2025 - Whisper Finetuning

Imaging Oxygen Dips in Mouse Cortex: Semi-NMF for Signal Extraction