Full-stack engineer working on AI and workflow automation. I like agentic systems with hard rules around them,
clean APIs, and interfaces people actually enjoy using.
- Now · Working Student Full-Stack Engineer, Workflow Automation & AI, at Zurich Insurance Group: AI-powered automation across the stack.
- Before · Machine Learning Engineer Intern at Earkick AI: multimodal (audio and video) emotion recognition.
- Wins · 🏆 Best Design, H0: Hack the Zero Stack (Vercel v0 & AWS) · 🥈 2nd Place, Dynatrace track, Google Cloud Rapid Agent Hackathon.
- Open source · A merged production fix in genblaze, a pull request to Amazon's AgentCore SDK, and vega-calm-ui on npm.
Demo video · Code · Project page
HSLU Exam Assistant: backend · frontend · API docs | Let Us Connect: backend · frontend
| Project | What I did | Status |
|---|---|---|
| genblaze · Backblaze Labs | Root-caused a production bug in genblaze-openai, where HTTPS reference images broke OpenAI image-edit requests (#253), and fixed it (#256). |
|
| bedrock-agentcore-sdk-typescript · AWS | AgentCore Memory session storage for Strands TypeScript agents: 60 unit tests, integration tests against a real AgentCore Memory. | |
| vega-calm-ui · my library | Accessible 10-foot UI components for Fire TV (Vega OS) that refuse to render below 32 pt or 7:1 contrast, plus an ESLint plugin. | |
| react-native-multi-tv-app-sample · Amazon | High contrast, larger text and reduce motion for Amazon's multi-platform TV template. |
Thanks for stopping by. Explore the repositories, or say hello.



