I'm a junior at Duke interested in the design of low-latency, high-throughput backend systems. I have experience working on ETL pipelines at growth stage startups and I enjoy thinking about algorithmic problems.
Right now, I'm on the Data Platform team at Ramp working on their internal feature store for ML decisioning models.
In my free time, I like to explore new tech, play with fire, and try different restaurants.
Feel free to reach out at aaronhsu4606@gmail.com.
A conversation-based diary using LLM agents with RAG-based memory store. Implemented vector-based memory system using pgvector and embedding similarity. Deployed with CI/CD pipeline using GitHub Actions and Docker on AWS EC2.
A streaming platform managed with Kubernetes with NGINX-based RTMP ingestion and HLS transcoding. Implemented distributed edge caching layer with configurable TTL and round-robin load balancing. Deployed with Grafana monitoring for cache performance optimization.