About
Hi, it's Xiangyi Li! Incoming intern at Tesla
Work Experiences
Founding Engineer at Imaginix
San Jose, US
Built the majority of the Kimi app, including the frontend, backend, database, logging, AI engineering, and CI/CD. Built the WebView layer for the Live2D model. Hand-crafted a suite of interview questions used to interview >25 people
Software Engineer Intern at Dolby Laboratories
Shenzhen, China
Developed Windows SDK using .NET framework for Dolby Vision color profile generation for >200 manufacturers and designers producing >500k laptops annually. Implemented RBAC for the support web portal and improved customer experiences with ASP.NET Core, SignalR, jQuery, and CSS. Actively collaborated with cross functional teams. Actively participated in the software development lifecycle using AWS and IIS.
Software Engineer Intern at Red Hat
Beijing, China
Extended Kubernetes APIs for automation in Golang for two OpenShift releases. Built OpenShift Dynamic Plugins using React and TypeScript for better feature visibility. Developed Cypress scripts for the OpenShift web console. Actively participated in debugging Jenkins CI, fixing Ruby testing scripts, and setting up AWS network environments.
Research Assistant at CUHK
Shenzhen, China
Implemented statistical algorithms like gradient descent and canonical correlation analysis in C++ and Python. Compiled Python code for auto-vectorization for improved computation speed. Distributed VMs for matrix calculations.
Education
Master's Degree in Computer Science
San Jose State University
Attended TechCrunch Disrupt 2023, Stanford Law x LLM Hackathon in Sep. Attending AI Engineer Summits and Next Conf this October!
Bachelor's Degree in Computer Science
Chinese University of Hong Kong
Participated in a campus startup Renter as a full-stack developer. Built a WeChat mini-app using Next.js, Node.js and MySQL.
Projects
Better Diffusion
Work in progress
Scraped, cleaned, and analyzed with NLP techniques on Midjourney's Discord messages. Used them for fine-tuning GPT-3.5. Generate prompts with OpenAI Function API and then images using stable diffusion.
Ask Abe AI
https://ask-abe-beta.vercel.app/
Developed a better law answering agents. Scraped law data with TypeScript and generated vector embedding with OpenAI embedding API. Used Supabase and pgvector with HNSW for RAG. Built the frontend using React Server Components and TypeScript.
PadCoder
https://github.com/l1xiangyi/padcoder/
Built an iPad LeetCode grinding app using React Native and TypeScript with another engineer. Participated in designing using Figma. Sold to an early stage startup by my partner.
VOC2012 Object Detection Task
http://github.com/l1xiangyi/dda4220-final/
Analyzed images with classic OpenCV algorithms. Fine-tuned STOA models like Faster RCNN, SSD, and YOLO and compared performances. Wrote Bash scripts to train and log results for GPU clusters.
ChatGPT Detector
https://github.com/l1xiangyi/csc3160-final/
Fine-tuned RoBERTa models with scraped human texts and generated texts using a suite of different LLMs. Compared the results with the baseline model. Proved increased generalization performance by introducting more LLM texts.
Kanan Minimal
https://trello-minimal-csc4001.web.app/
A Kanban task management web app built with Angular, Nest.js, and Firebase for the Software Engineering course.