Technical Background
I work across agent systems, model evaluation, and product engineering. The projects below show how I use those capabilities in practice.
Agent systems
Tool use, memory, task decomposition, document processing, structured outputs, and human-in-the-loop review. My focus is the path from a user request to artifacts that can be inspected and used within an existing workflow.
In practice: Nova Agent, TikTok coding agent, and Infant Agent.
Evaluation and model research
Code-generation evaluation harnesses, prompt iteration, instruction-driven editing, and structure-aware generation. My research also covers real-time inference under changing environmental conditions.
In practice: FineEdit, TreeDiff, and RED.
AI product interfaces
Full-stack applications, mobile interfaces, streaming interactions, and operator review flows. I use prototypes to explore what users need to see and control when working with AI.
In practice: AgentShelf, Obi, and Live Subtitles.
Engineering foundation
| Area | Tools and experience |
|---|---|
| Languages | Python, TypeScript, JavaScript, SQL, Swift, Objective-C, Java, C++ |
| Model development | PyTorch, TensorFlow; earlier computer vision and GAN research |
| Product development | React, Next.js, React Native, SwiftUI, Vue |
| Services and data | Node.js, Python backend frameworks, REST APIs, PostgreSQL, MySQL, Redis |
| Delivery | Docker, AWS, Vercel, GitHub Actions, Linux, Git |
Earlier roles at Ant Group, Ansys, and Pittsburgh developed my experience with production services, deployment workflows, and software for research teams.