Experience

A curated snapshot of my professional experience and what I shipped.

Software Engineer Intern

Riot Platforms, Inc. • Austin, TX • Intern

2025-06-03 — 2025-08-10

Designed an AI agent in C# & .NET 8 with Semantic Kernel, deployed on Amazon EC2.
Supports natural-language queries on AWS Timestream data, with MCP integration for Bitcoin mining ops.
Engineered a hybrid RAG search (BM25 + dense vectors) for Bitcoin mining ops Q&A.
Deployed an Agentic workflow with GPT-4o for planning and Claude 4 Sonnet for complex SQL generation.
Refined the agent via iterative prompt tuning, expanding accurate query coverage by 10x.
Delivered 90% higher query accuracy, 70% faster response time, and 80% lower operational cost.

C#.NETSemantic KernelAWSAI Agent

C-STAR • New York City, NY • Intern

2024-10-01 — 2026-03-01

Built a web app with FastAPI, React, and MySQL, reducing manual real estate processes by 95%.
Engineered data pipeline using AWS S3, EC2, and Lambda for automated batch ingestion.
Deployed SageMaker with Random Forest Regressor to predict property prices with 92% accuracy.
Implemented real-time data sync between MySQL and Google Sheets API for dynamic property listings.
Created rolling storage solution to reduce Google Sheets storage usage by 90% while preserving data integrity

PythonFastAPIMLOPsAWSMySQLReact

University of Science and Technology Beijing • Beijing, China

2023-07-01 — 2024-07-01

Proposed utilizing Transformer to model the spatio-temporal relationship between human posture and clothing movement
locally and globally, considering that humans wearing loose clothing were difficult to model by NeRF (Neural Radiance
Fields) under the guidance of the SMPL (Skinned Multi-Person Linear) model.
Generated more realistic dynamic details of clothing and handled the movement of clothing (e.g., open jackets), compared
to methods that rely heavily on the nude body template topology. Enhanced training efficiency by 40% through the
implementation of Distributed Data Parallel (DDP) for multi-GPU training,
Developed a professional Web https://github.com/MatrixBrain/awesome-NeRF and gave back to the research area by
cataloguing NeRF’s papers, which was highly recommended by other researchers.