Projects

Graph Learning

• Torch-Quiver: A Scalable Graph Learning System with Adaptive End-to-end Parallelism

  Scalable training of graph neural networks

• DGL-KE: Training Knowledge Graph Embeddings at Scale

  Distributed knowledge graph embeddings

Databases

• PD: Placement driver for TiKV

  Making data paritition and replication adaptive to the query workload

• Fugue: Distributed Databases with Decoupled Transaction Services

  Decoupling transactions and data analytics to build a real-time HTAP system

• S3: A Scalable In-memory Skip-List Index for Key-Value Store

  Optimizing the index structure for LSM-tree engines