Di Chen Email & Phone Number

Cambridge, Massachusetts, Us

Responsible for building the ML powered sales platform at Hubspot.

Brooklyn, Ny, Us

- Led the development of Model Registry within the ML Platform, centralizing the storage of model artifacts and metadata with version control. Led the effort of onboarding 100+ models to the registry, significantly improved model accessibility and tracking at Etsy.- Led the integration of Model Registry with ML systems at Etsy, encompassing training, offline experiment, profiling and serving. Developed a Python SDK to facilitate streamlined model registration and deployment process, reducing the manual effort in the model development lifecycle by 80%.

San Jose, Ca, Us

Led the development of eBay's unified inference platform on Kubernetes, deployed across multiple data centers. This platform supports 200+ models and serves 30 billion daily inference calls, ensuring high availability and fault tolerance with multi-cluster support and DNS-based Geolocation Routing. • Designed and implemented eBay unified AI inferencing platform. - Built on eBay in-house Kubernetes cluster, powers high throughput model inferencing with low latency, high availability and fault tolerance. - Spearheaded the enhancement of online prediction capabilities by integrating advanced inferencing patterns like champion/challenger and preprocessing, and query-based routing. - Developed the self-service model CI/CD pipeline using Jenkins and Kubernetes API, enabling automated model verification and streamlined deployment process.• Designed and implemented eBay Core AI SDK for Python. - Implemented model registry APIs for model discovery and model sharing. - Integrated with notebook to support interactive model development for serving. - Integrated with feature store to support feature discovery and offline feature publish. - Implemented client APIs for model logging, monitoring and validation.

San Jose, Ca, Us

- Developed marketing email triggers within eBay's Targeting Engine, including Purchase Journey, Daily Trigger, and Event Digest. This initiative powered approximately 50 million daily email sends and contributed to 100 million annual GMV.- Contributed to the marketing near real-time data pipeline, processing 50 million user activities and 2 billion item listings with minute-level delay to Couchbase, bolstering the effectiveness of marketing ML models with up-to-date online features.- Redesigned the Targeting Engine using Spark workflows, significantly improving efficiency with a 70% reduction in batch execution time and enabling send-time rendering for marketing emails.

Beijing, Cn

Core member of the Hadoop storage team, managing the large scale HDFS cluster with over 20000 nodes and handling more than 100PB of data across the company. Key achievements include:- Led the master node availability optimization by implementing a hot standby mechanism for automatic failover and seamless switching, resulting in a 90% reduction in downtime during master node failures. - Led the Namenode Federation project, amplifying inode capacity by 10x and providing dedicated Namenode for domain-specific resource isolation.- Implemented the transparent compression strategy for Datanode, enabling background compression of cold data blocks, thereby reducing disk utilization by 30% across the platform.

Shenzhen, Guangdong, Cn

• Implemented the query module for merchants and charge system of the Tencent Online Shopping Platform.

Master'S Degree, Computer Science

Bachelor'S Degree, Computer Science