Wang Zuo Email & Phone Number

Shanghai City, China

Wanda AI Lab is reorganized as DeepInsight AI Lab https://github.com/deepinsight1.Apply BigData, Cyber Security, and BlockChain to Internet finance, investment, M&A and insurance, providing one-stop innovative financial services to retailers and consumers in Wanda's offline commercial platform - the largest of its kind in the world. Supporting Wanda Finance Group's relentless expansion in micro-loans, credit scoring, insurance and investment etc.2.Architecture on the state-of-art database management(OLTP) and data warehouse(OLAP) to provide fast analytic, full ACID, decentralized systems. Build high concurrency blockchain platform based on Hyperledger, provide pluggable db and consensus. 3.Design algorithms for various industrial applications for machine learning, including recommend system and anomaly detection.

Shanghai City, China

1. Optimize ApacheImpala with AVX2: vectorized query execution, batch operates of bitmap filter. The speedup for bitmap operators is 7.1X(get), 4.0X(or), 3.5X(and), 5.6X(set). Invented new compression technology for Parquet format file and new ways to evaluate predicates on encoded data. The speedup is 2.6X ~ 3.8X for compression and 6X ~ 19X for decompression. The speedup for TPC-H queries is 2.14X (q6), 1.47X (q13), 2.0X(q14), 1.45X(q15), 1.32X(q20). Co-presented the work at Strata Hadoop NYC. Implemented ByteSwap with AVX2, the speedup is 10.7X ~ 38.6X. Led the investigation, algorithm design, and benchmark in AVX sort project. Invented a new approach to encode integer into double so that merge times reduced to at most once. With this new approach, the speedup of AVX sort is more than 3X.2.Involvement and leadership in projects: In-depth investigated into cache-aware hash join, Crystal Ridge memory, parallelism on multi-cores, lock elision with RTM and HLE. write technique reports and proposals.3.Contribute to patches to Apache Impala: Patches relates to UDF, compiler, security, frontend and backend of Impala.4.Research on the state-of-art database technologies: Morsel-Driven Parallelism, seastar framework(for high performance OLTP work), RDMA

Hangzhou, Zhejiang, China

Work as a designer and mentor in SQL Engine team. SQL Engine can generate SQL from a report or document and analysis data in in-memory cube. Design and implement features in C++. Conduct research on big data engine, analysis performance for MapReduce, HBase, Impala, Solr, Greenplumdb. Hack on HBase’s coprocessor, and Solr’s analytical component. Design and implement algorithm for relation generation and in-memory cube join. The challenge for relation generation is analyzing linked graph, and global optimize graph, and for cube join is to analyzing tree built by decomposing grid elements. Nielsen project: metric threshold, filter with dimensionality, derived elementIn-memory DB project: Multi Cude Join Engine, Direct Data Access, Derived Attribute