Thanos
Company
Cowlar Design Studio
Role
Full Stack Engineer
Timeline
2023 - 2024
Links
Live cart footage streamed with synced AI prediction overlays
Leaderboards built to speed up review throughput
Dockerized microservices scaled to review workloads
In short
Thanos is a human-in-the-loop review platform: a real-time video player interface streaming live cart footage, with AI predictions overlaid so annotators can review and correct them on the spot. I built the platform — frontend, backend APIs, and database; the AI/ML model itself was built by a separate team.
The problem
Annotators needed to watch live video streams of carts, see the AI predictions for each, and correct them in real time, but no tooling existed for that workflow.
What I did
Built dynamic browser-based video review interfaces in Vue.js, with a real-time player showing live prediction overlays annotators could accept or correct. Built the backend APIs and database in Node.js and MongoDB, plus a gamified leaderboard dashboard to keep annotator throughput high.
The result
- Built a working real-time review workflow for live video and AI predictions
- Gamified leaderboards to speed up annotator throughput
- Deployed and scaled on Kubernetes
Technical detailsFor engineers: how it's built and what it runs on
Architecture
- Vue.js frontend with a real-time video player and prediction overlay UI
- Node.js/Express backend APIs and MongoDB for review and leaderboard data
- Live streaming pipeline for cart footage with synced prediction data
- Dockerized microservices deployed on Kubernetes
Highlights
- Real-time video player streaming live cart footage
- Live AI-prediction overlays annotators can review and correct
- Track annotator performance in real time
- Gamified leaderboards to boost productivity
- Collaborative platform for annotation teams
- Auto-scaling for any project size
- Fast setup and deployment from day one
- Handles large video and annotation datasets at scale
Stack
Finished reviewing?
Back to Portfolio←