Xinghan Yan

Computer Science Graduate Student at USC

I'm a Computer Science graduate student at USC pursuing a Master's in Computer Science with a focus on Game Development. I have extensive full-stack development experience, strong problem-solving skills, and the ability to effectively communicate and implement technical requirements. I am well-versed in code architecture and algorithms, and have rich experience in teamwork and cross-functional collaboration. I am seeking a position related to Software Development Engineering.

Education:
University of Southern California | Master of Science in Computer Science - Game Development | Aug 2025 -- May 2027
Northeastern University | Bachelor of Science in Computer Science | GPA: 3.61/4.0 | Sep 2021 -- May 2025
Khoury College of Computer Sciences

Technical Skills

Languages: Java, Python, JavaScript/TypeScript, C/C++/C#, SQL, HTML/CSS

Frameworks & Libraries: Spring Boot, React, Node.js, Express, Flask, Jest, Apollo, WebSocket, gRPC

Data & Storage: MongoDB, MySQL, Redis, Kafka, Flink, Vector DBs

Machine Learning/AI: PyTorch, TensorFlow, DeepFM, DSSM, PPO, DDPG, TD3, Contextual Bandits

Cloud & DevOps: Docker, Kubernetes, AWS, CI/CD, Monitoring/Logging

Tools: Git, Jira, Postman, Figma, Unix/Linux Shell, LaTeX

Working Experience

OAG Technology | Data Scientist Intern - Smart Bidding | May 2025 -- October 2025 | New York, NY

• Built industrial-grade RTB engine with Flink/Kafka/Redis, sustaining 10W+ QPS and 20–50ms E2E latency for real-time bidding at scale.
• Engineered real-time feature pipelines for user embeddings (DeepFM/DSSM) and CTR/CVR feedback, achieving sub-50ms feature computation.
• Designed bid strategy models combining CTR/CVR estimation with bid landscape modeling and eCPM calculation.
• Implemented RL-driven bidding strategies: Contextual Bandits (LinUCB, NeuralUCB) for cold-start users and deep RL (PPO, DDPG, TD3) for budget pacing.
• Achieved 18.5% ROI uplift, 10% lower budget consumption, and +12% CTR lift through constrained RL and multi-agent coordination.

Li Auto | Application System Development Engineer | July 2023 -- December 2023 | Beijing, China

• Led backend code refactoring initiative in Java/Spring Boot, restructuring business logic layer into clear controller/service/repository components. Reduced data misclassification incidents by 40% and improved platform efficiency by 25%.
• Designed and implemented REST API ingestion for nationwide competitor sales data with input validation, retry policies, and batch upserts into MySQL. Enhanced schema with normalization and composite indexes, sustaining 99.9% accuracy and improving uptime from 95% to 99.5%.

Projects

HuskyFlow (TypeScript, Node.js, MongoDB, OpenAI, WebSockets) | April 2025

• Developed full-stack Q&A with React/Node/Express/MongoDB, JWT authentication, RBAC, and real-time chat using WebSockets.
• Integrated OpenAI API via backend orchestration with request shaping, moderation, and load management for scalable AI answers.
• Implemented Semantic Search and Personalized Recommendations via OpenAI's text-embedding-ada-002 and cosine similarity for enhanced content discovery, structured for vector DB and RAG-ready retrieval.
• Live site

Kambaz LMS + Piazza Q&A Platform (React, TypeScript, Node.js, MongoDB) | March 2025

• Built a full-stack LMS with course/module/assignment management and role-based permissions.
• Integrated a Piazza-style Q&A forum with threaded discussions and visibility controls.
• Live site

Custom Reliable Transport Protocol (Python, UDP, Sockets) | Oct 2024

• Implemented reliable UDP protocol with sliding window, retransmission, and ACK/dup suppression for in-order delivery.
• Engineered sender/receiver with raw sockets; stress-tested under packet loss, duplication, reordering, and delay.
• Tuned timeout and window size for throughput, validated with telemetry; achieved lowest bytes sent and 2nd fastest runtime.
• Github