Ziqiao Xi - Resume

Work Experience

Quantum Compliance May 2024 – August 2024

Full Stack Developer Intern - Ann Arbor, MI

Developed an end-to-end solution for parsing and extracting data from chemical safety data sheets using GPU-accelerated OCR (Tesseract) and clustering techniques, reducing processing time from 15 to 3 minutes per PDF.
Deployed locally hosted LLMs (Mistral, Llama3 via Ollama) to ensure data privacy, orchestrated queries using LangChain, and integrated selective OpenAI API usage for enhanced accuracy.
Implemented and optimized RabbitMQ message queues to distribute PDF parsing tasks across GPU-backed servers, incorporating retry logic and dead letter queues to handle failures and analyze problematic messages for reliability.
Implemented an Excel export feature in a Spring Boot and Angular application, integrating a seamless UI button and leveraging Apache POI to generate well-structured spreadsheets for tracking workers' certificate statuses.
Optimized legacy SQL queries by resolving the N+1 problem, consolidating multiple inefficient loops into a single query strategy, reducing data retrieval and file generation time from 1 minute to 5 seconds
Deployed the service on GPU-backed servers using Docker containers and designed Nginx load balancers to ensure efficient distribution and scalability.
Designed and implemented a Retrieval-Augmented Generation (RAG) chatbot using LangChain, incorporating a sliding window technique to deliver real-time, context-aware answers to chemical regulation queries.
Scraped chemical regulation websites using Python, preprocessed data into multilingual vector spaces with Hugging Face's GPT-2 tokenizer, and employed Faiss for vector storage, similarity search, and automated index updates.

Project

Meditation App www.ziqiaoxi.com

Independent Full Stack Developer

Developed a scalable meditation-sharing platform where users can record, post, and share meditation experiences with text, images, and location data using Node.js and React.
Designed an optimized MySQL database with ER diagrams for user posts, comments, and location data, implementing indexing and query optimization techniques to improve retrieval speed.
Optimized MySQL query performance by implementing Redis as an in-memory cache with LRU eviction policies, fine-tuned key design, and expiration strategies to efficiently cache user data and session tokens for faster access.
Built scalable RESTful APIs using Node.js and Express.js to handle CRUD operations on MySQL, with pagination implemented for efficient retrieval of large datasets, such as user posts and comments.
Integrated RabbitMQ as a message broker for tasks like asynchronous image uploads and processing, enabling efficient handling of large image files while decoupling services for better maintainability and scalability.
Designed Elasticsearch indexes using an inverted index structure for efficient text and location-based search, enabling fast and accurate retrieval of user posts based on keywords or proximity to specific locations
Configured Prometheus for real-time monitoring of system metrics, setting up Slack-based anomaly alerting.
Deployed the app on Dockerized EC2 instances with Kubernetes; implemented GitHub Actions and Argo CD for CI/CD, automating unit and integration tests with Jest/Supertest and multi-architecture Docker build.

Operating Systems January 2024 - May 2024

Course Project - Ann Arbor, MI

Built a custom user-level thread library using C++, supporting context switching, thread creation, and FIFO-based scheduling with safe concurrency through interrupt management and global atomic guards.
Developed a demand-paged virtual memory manager with swap/file-backed pages, clock-based page replacement, copy-on-write optimizations, and a robust state machine for efficient page lifecycle management.
Designed a secure, multi-threaded network file server with password-based encryption, an upgradeable reader-writer lock scheme, and a non-hierarchical file system structure to ensure crash safety and file system consistency.

Disributed System January 2024 - May 2024

Course Project - Ann Arbor, MI

Designed and implemented a distributed MapReduce framework with a Manager and Workers communicating over TCP/UDP, featuring job queuing, task scheduling, and dynamic worker registration.
Implemented a concurrency model using upgradeable reader-writer locks to synchronize threads, prevent data races, and manage job queues, worker states, and task assignments.
Optimized performance with a streaming approach for mapper and reducer input/output,,and incorporating fault tolerance via UDP heartbeat monitoring to detect Worker failures and reassign tasks seamlessly.

Web Search Engine January 2024 - May 2024

Course Project - Ann Arbor, MI

Developed a web search engine using Flask with a multi-stage MapReduce pipeline, Index and Search servers, and a frontend interface for ranking and displaying query results using React
Built an inverted index using a MapReduce pipeline to compute tf-idf scores, segmented indices by document ID, and integrated PageRank data to enhance document ranking with user-configurable weight adjustments.
Developed a RESTful API in the Index server to retrieve query results as JSON, optimized query processing by leveraging precomputed index segments stored in memory.
Incorporated upgradeable reader-writer locks to ensure thread-safe operations in the Index and Search servers, enabling concurrent read operations and writes for in-memory updates and query result merging.

Education

Work Experience

Project