01 / Human–AI
Knowledge Bridge
I joined an early knowledge-graph proof of concept and refactored its architecture, performance, UX, and human–AI interactions.
81% less loading time · 72% faster task completion
- Contribution
- Full-stack development, UX, performance, human–AI interaction
- Focus & technologies
- React · TypeScript · WebSockets · Python · FastAPI
Starting with a proof of concept
The graph-generation work already existed. My contribution focused on making a broken proof of concept usable, responsive, and easier to explore.
Progressive, on both sides
WebSocket-based progressive transmission, backend graph pagination and prioritization, and custom viewport- and zoom-dependent lazy rendering reduced the amount of information arriving and rendering at once.
Keeping people in the loop
People could correct AI-generated tags and relationships. Those corrections were sent back to the backend to support learning. The work achieved 81% less loading time and 72% faster task completion.