Hi, I'm
Emmeline.
Software Engineer shipping LLM-powered products at Readgates. I architect the kind of systems that handle millions of requests, reason over documents in milliseconds, and collaborate in real-time — without breaking a sweat.
The person behind the code
Full-stack engineer with a deep interest in distributed systems and AI infrastructure.
I'm Yiren (Emmeline) Xu — a Software Engineer at a startup, where I spend my days building RAG pipelines, designing real-time collaboration systems, and making LLM inference actually fast. It's the kind of work that sits at the intersection of AI and backend engineering, and honestly, I wouldn't have it any other way.
I completed my Master's in Software Engineering at UC Irvine in 2025, where I got obsessive about distributed systems, concurrent programming, and software architecture. Before that, a dual BS in Computer Science & Accounting at Kean University — yes, the accounting degree comes in handy when pitching the ROI of a refactor to stakeholders.
I care about code that's readable, systems that are observable, and products that people actually enjoy using. Outside of engineering: miHoYo (GenShin, 3Z, Star Rail), Reading, Cooking.
UC Irvine · 2025
Where I've built things
From AI-powered platforms to healthcare apps to urban data systems.
- Built an LLM-powered RAG pipeline with document chunking, vector embeddings, and MongoDB sharding — enabling semantic search across user docs at ~200ms retrieval latency.
- Designed a real-time collaboration layer using WebSocket + event-driven state management, enabling AI-assisted co-editing and cutting frontend latency by 20%.
- Built a distributed media processing pipeline via Kafka Streams + AWS Lambda — async processing, live progress tracking, 99.5% task reliability.
- Deployed LLM inference services with Ray-style pipeline orchestration and request batching to scale AI features without blowing up costs.
- Built a scalable ETL pipeline with Apache Flink + Snowflake for energy data processing; materialized views + partitioning delivered a 3× query performance boost.
- Shipped an NLP service backed by the Gemini API + Redis Cluster caching, hitting sub-200ms latency under 50+ concurrent requests.
- Implemented full observability (Prometheus + Grafana) across the stack, deployed on Azure VM — 99.9% uptime, zero guessing when something breaks.
- Built a healthcare app in React Native + GraphQL/Apollo with an AI chat interface and real-time appointment updates via WebSocket.
- Designed a two-tier caching system (Redis Cluster + in-memory local cache) for microservices, dramatically cutting redundant DB queries.
- Implemented distributed transaction management using Kafka + local message table; PostgreSQL schemas with table partitioning to handle appointment scheduling at scale.
- Developed Spring Boot microservices for a city management platform handling geospatial data — Redis caching + spatial indexing pushed it to 500 QPS.
- Built RESTful APIs with MyBatis for complex geospatial ops; custom indexing + caching strategies trimmed read latency by 30%.
Things I've shipped for fun
Hackathons, side projects, and experiments in AI + systems.
- LLM-powered ingredient extraction feeding a lightweight RAG pipeline with vector search — the fridge talks, the model listens.
- Agentic workflow where an LLM orchestrates ingredient parsing, recipe retrieval, and ranking to generate step-by-step cooking reasoning.
- Data pipeline for structuring recipe datasets and generating embeddings for semantic search across 10k+ recipes.
- LLM inference APIs deployed on AWS EC2 + Vercel with request batching and streaming responses for snappy UX.
The toolkit
Languages, frameworks, infrastructure — the full stack, literally.
Where I learned to think
Two degrees, one obsession: building systems that actually work.
Donald Bren School of Information & Computer Sciences
College of Science, Mathematics and Technology
Let's build something
Whether it's a full-time role, a project collaboration, or just a good conversation about distributed systems — my inbox is open. I reply fast.