Designed and developed high-availability payment gateways and core transaction routing systems for Alipay US using Java and Microservices architecture, supporting massive-scale financial transactions.

Tackled complex distributed system challenges, including distributed transaction management, API idempotency, and strong data consistency in high-concurrency environments.

Optimized the performance of core payment processing flows by implementing distributed caching and asynchronous message queues, significantly improving system throughput (TPS) and reducing latency.

Integrated local financial institutions and payment channels in the US market, designing flexible payment flows while ensuring the underlying architecture complied with strict regional financial and data security standards.

Developed and incorporated scripts using Python, Google Cloud Platform, and AWS to build a pipeline for data processing, to improve the efficiency of preparing training data.

Leveraged PyTorch, and applied Transfer Learning paradigm to experiment with VGG-19 and ResNet-50 to train a CNN model to recognize the medium class of an artwork.

Implemented a patch-based CNN using PyTorch to recognize the artworks of an artist, scaled the model to multi-artist classification, discovered the theoretical limitation, and optimized the data processing to improve the performance.

Deployed deep learning models as a web app using Flask, and built REST APIs for model services using Flask RESTful to interact with other applications online and make the models act on time when they are called.