• HFT System

• Designed and developed a Multi-Armed Bandit system for personalized item recommendations.• Created pipelines for text-based and collaborative filtering-based recommendations to expand the item pool.• Implemented draw and reward modules using Java and Python.Spark, SQL, PyTorch, AWS

• Built a data pipeline entirely from scratch to effectively analyze important business metrics.• Developed business APIs and automated CI/CD.Rails, AWS, Kubernetes

• Built a face recognition model for TalkDrawer service, similar to Google Photos. Conducted experiments to continually enhance the model's performance.• Employed a range of techniques such as GAN and active learning to collect and refine face recognition data.• Developed pre-processing and post-processing modules using Python API and Triton server to serve the model.PyTorch, Kubernetes

• Engineered a distributed search cluster with integrated business logic to deliver reliable search results at scale.• Designed and implemented an indexing document generator that extracts information from multiple domains.• Established an address search engine from the ground up.• Constructed a custom solr plugin framework for managing various ranking signals.• Maintained the product search system, which included solr, solr components, a query broker, a document generator, a java adapter library, and an HTTP API.Solr, Java, SQL, AWS