Yiwei Zha

Background

About Yiwei Zha

Software Engineer & AI Researcher Hi! I’m Jonathan Zha, a software engineer and researcher passionate about building high-throughput distributed systems and advancing AI for scientific applications. I specialize in solving problems where technical rigor meets real-world impact—whether optimizing cloud-native architectures or designing physics-aware ML/DL models. What I Bring to the Table Scalable Systems: Built Java/AWS microservices (Upic Ski System) handling 2,500 requests/second using dynamic thread pooling and RabbitMQ for decoupled messaging, while designing BI dashboards (ByteDance) with React/Spring Boot to visualize 1M+ daily game ad metrics, reducing load time by 35% through server-side pagination. Cloud Architecture: Deployed AWS EC2-powered systems (Upic) with containerized scaling to achieve 40% throughput gains, validated via JMeter load testing, and architected FastAPI (ByteDance) with Celery/NGINX to handle 500+ concurrent ad bid requests, improving throughput by 60%. Automated monitoring with Prometheus/Grafana (ByteDance) to detect slow queries and pricing anomalies, reducing latency by 40% and financial errors by 80%, while ensuring zero-downtime microservices (Upic) via JSP 6.0-based skier/resort APIs. AI for Science: Developed physics-informed neural networks (PINNs) for lithium-ion battery health estimation, integrating electrochemical PDEs with LSTM models to improve SOH prediction accuracy by 22%(published at NU IMPACT Symposium). LLM Innovation: Co-created MobileAgentBench, a benchmark for mobile LLM agents, leveraging Android emulation and multi-modal task pipelines to expose 35% accuracy gaps in real-world mobile interactions (Accepted at 2025 MARW workshop and nominated for JMLR). Recent Highlight Presented groundbreaking research on Physics-Aware Battery Health Estimation at the NU IMPACT Symposium 2023, earning Best Poster Award for demonstrating a 40% error reduction in predicting lithium-ion battery lifespan. This work bridges AI with material science, offering tangible applications for EV sustainability. Let’s Connect If You’re Passionate About Architecting distributed systems Advancing AI for Science Solving high-stakes engineering challenges with metrics-driven precision Open to roles in SDE or AI research. Let’s discuss how we can push boundaries—in code, collaboration, or clean energy! 🤖