Projects

A public catalog of more than 4,000 real enterprise AI deployments, each traced to a public source, with insights and trend analysis on top.

Crypto signals, daily model rankings, market research, and paper portfolios, built on XGBoost ranking models and walk-forward validation.

Read-only Hyperliquid trading analytics: paste an address and get an honest, data-backed breakdown of how that account trades.

A local-first workbench for measuring how well voice models understand Swiss German dialects, with WER, CER, chrF, and streaming latency.

An in-progress robotics research simulator for embodied agents, with a React operator cockpit, agent-guided navigation, sensor telemetry, and reproducible test scenarios.

An open-source agentic pipeline for discovering, extracting, deduplicating, and reviewing structured evidence from public websites. It powers AIUseCaseHub.

A Microsoft Agent Framework and Foundry demo that turns a customer opportunity into a cited, enterprise-ready solution proposal, visualized as a pixel-art agent office.

A hosted MCP server that lets agents search source-linked enterprise AI deployments and retrieve the evidence behind them.

Turns real bug-fix pull requests into reproducible, reward-bearing tasks for training and evaluating coding agents.

A technical AI blog I started as a machine learning consultant, with hands-on tutorials on Python, machine learning, RAG, AI agents, and real-world enterprise AI use cases.

A legal AI agent for Swiss law that supports multiple case types and works across cases, regulations, and related legal sources.