Tianyi Lu Email and Phone Number

- Working as a lab assistant and grader for CS208 computer system course

- Developed a cross-platform stock portfolio monitoring and management application from scratch using Flutter. Maintained connections with multiple gRPC endpoints using protobuf, and enhanced application performance by implementing memorized requests, utilizing Riverpod state management paradigm and optimizing asynchronous data handling, resulting in a 40% increase in data processing speed and UI responsiveness.- Developed an optimized data processing pipeline to generate and store daily statistics for 60GB of Chinese stock market data. Reduced memory and disk usage through the efficient use of C struct binary files and SQLite. Enhanced data processing speed by utilizing vectorized operations in Pandas.

- Mirro AI is an LLM-powered community building application dedicated to establishing stronger and long-lasting personal connections. We were admitted to MiraclePlus (previous Y-Combinator China) S23.- Developed the main web application using React.js with TypeScript as one of the four founding engineers. Engineered instant messaging, contact and group chat management, and picture-sharing plaza, all integrated with innovative GPT usages as core features for our product. Utilized Tailwind CSS and Framer Motion to create a dynamic and interactive UI.- Combined client-side data caching with SWR and server-side rendering using Next.js to achieve under 0.1s average response time. Collaborated with mobile developers to compile the web UI into cross-platform apps using Capacitor.

- Generated interactive speaking avatars for elementary school students by integrating generative AI tools (ChatGPT, Stable Diffusion, and D-ID) with a React.js web UI and type-safe REST APIs using Python FastAPI and Pydantic.- Utilized Async IO to handle concurrent requests and deployed an IP hashing load balancer to manage GPT endpoints. Implemented retry logic in GPT requests with response caching in SQL database with SQLAlchemy and SQLModel ORM to reduce external API costs by 70%.- Engineered a GPT-powered chat survey platform to convert open-ended responses into quantitative data, which was then used to design an assignment system for 500+ students and faculties, integrating GPT-driven analytics with stable matching algorithms.