Horjun Streaming Platform (Go, Next.js, Kafka, ClickHouse, FAISS)
Developed a high-performance, microservice-based video streaming architecture designed for low-latency delivery and personalized user experiences. The system was built to handle high-concurrency traffic and massive datasets, leveraging a modern backend stack for data integrity and real-time processing.
-
Core Technologies:
- Backend: Go (go-zero framework)
- Frontend: Next.js, Material UI, Redux Toolkit
- Databases: PostgreSQL, Redis, ClickHouse
- Infrastructure: Kafka, MinIO, Meilisearch
- AI/ML: FAISS (Facebook AI Similarity Search)
-
Key Technical Contributions:
Backend & Infrastructure (Golang Microservices)
- Engineered the backend using the go-zero framework, implementing robust service discovery and internal RPC communication for high availability
- Managed a polyglot persistence layer: PostgreSQL for relational metadata, Redis for caching, ClickHouse for analytical data
- Integrated Kafka for asynchronous tasks: view-count updates, log processing, notification triggers
- Utilized MinIO for scalable object storage and Meilisearch for sub-100ms full-text search
Recommendation Engine
- Implemented a vector-based recommendation system using FAISS for real-time “Similar Videos” suggestions based on content similarity
Frontend (Next.js)
- Built a responsive, SEO-optimized frontend with Material UI
- Managed global state (auth, playback preferences, themes) using Redux Toolkit
-
Impact/Results:
- Independent scaling of video-serving and auth services during peak traffic
- Reduced search latency to sub-100ms via Meilisearch and Redis caching
- Increased content discoverability and user retention through FAISS-driven similarity engine