Dr. Yongchao Liu is currently a Staff Engineer in Ant Group (from 07/2017 to present), working on distributed machine learning (online and graph learning) and heterogeneous computing platforms for deep learning (neural network compressors and compilers). Prior to that, he worked as a Research Scientist II (a research faculty member) in the School of Computational Science & Engineering, Georgia Institute of Technology (USA) (from 01/2015 to 07/2017), and as a Postdoctoral Researcher in the Institute of Computer Science, University of Mainz (Germany) (from 11/2011 to 01/2015). He earned his Ph.D degree in computer engineering from Nanyang Technological University (Singapore) in 2012 (supervised by Dr. Bertil Schmidt and Dr. Douglas Maskell). Prior to that, he earned the Master and Bachelor degrees in computer science and technology from Nankai University (China) in 2008 and 2005, respectively. He is a non-IEEE member and a non-ACM member.
He is an internationally recognized researcher in parallel computing and bioinformatics. He demonstrated outstanding contributions to scientific communities by his novel parallel algorithms and software tools for large-scale biological data analysis. These parallel algorithms and tools streamline fundamental and computationally challenging biological issues with parallel computing, via in-depth exploration of current high performance computing techniques and technologies such as hardware accelerators (e.g. GPUs and Intel Xeon Phis) and clusters (e.g. CPU/GPU/Xeon Phi clusters). On the other hand, these algorithms and tools investigate a set of related critical issues in bioinformatics and have actually established an innovative and unique analysis platform for large-scale biological datasets, especially next-generation sequencing reads. To address big data challenges, he proposed a new concept of computing, i.e. Compact Computing, which centers around data and enables full-stack computation by tightly coupling algorithms with systems. His research interests focus on parallel and distributed algorithm design for bioinformatics, heterogeneous computing with accelerators (GPUs and Xeon Phis), high performance computing on big data, and parallelized machine learning.
He has released a set of software tools for reproducible research and public use, most of which are open-source, associated with his paper publications. Among these algorithms, three CUDA-based open-source software tools, i.e. CUDASW++, mCUDA-MEME and CUSHAW, are rated by NVIDIA Corporation as popular GPU-accelerated applications, while DecGPU (the first parallel and distributed error correction algorithm for high-throughput short reads) was reported by GenomeWeb. Furthermore, some of his other open-source tools are also leading in their respective areas, including MSAProbs (multiple protein sequence alignment), Musket (Illumina reads error correction), CUSHAW2 (NGS base-space read alignment), CUSHAW3 (NGS base-space and color-space read alignment), PASHA (NGS de novo genome assembly), SNVSniffer (NGS germline and somatic SNV calling), SWAPHI (Xeon-Phi-based protein database search), SWAPHI-LS (Xeon-Phi-based pairwise DNA sequence alignment), ParaBWT (parallel construction of Burrows-Wheeler transform and suffix array), LightSpMV (CUDA-based sparse matrix-vector multiplication), and LightPCC (parallel correlation computation on Xeon Phi clusters).
He won two Best Paper Awards at IEEE International Conference on Application-specific Systems, Architectures and Processors (IEEE ASAP) in 2009 and 2015, the Program to Empower Partnerships with Industry Award from U.S. South Big Data Hub in 2016, Annual Excellent Patent Award of Ant Group in 2020, Annual Best Paper Award of Ant Group in 2021, and Annual SuperMa CEO Award of Ant Group in 2021. Besides, he got one Best Paper Award recommendation at IEEE International Conference on Cluster Computing (IEEE Cluster) in 2014. In Ant Group, he led his team to have broken the world record in the Stanford Open Graph Benchmark proteins dataset leaderboard in 2021, and to have developed a large-scale distributed graph learning system, named GeaLearning, based on a brand-new computational paradigm. The landing of GeaLearning on Zhima Credit in Alipay App was listed in Ant Technology Memorabilia in 2020. Moreover, GeaLearning is a core componenent of GeaGraph (alias TuGraph) that won the World's Leading Internet Scientific and Technological Achievement award from World Internet Conference in 2021. He was listed in the reputable Who's Who in America (Marquis research) in 2016. He was awarded Innovative Talent (in artificial intelligence) of Hangzhou 521 Program for Global Talents Introduction in 2019 and was recognized as Hangzhou High-Level Talents of Category C in 2020. In addition, he serves as a reviewer for some top journals, such as Nature Methods, Nature Communications, Bioinformatics, TPDS and TCBB, and as a program committee member for some leading conferences such as IPDPS and CCGrid.
- Out paper "TeGraph: a novel general-purpose temporal graph computing engine" has got accepted by ICDE 2022. TeGRAPH is the first general- purpose temporal graph computing engine for efficiently pro- cessing temporal path problems and their applications. This engine can achieve significant speedups over the state-of-the-art designs with up to two orders of magnitude (241×) with the throughput of two hundred million edges per second.
- Ant Group's large-scale graph computing system GeaGraph gets recognized as World's Leading Internet Scientific and Technological Achievement at World Intenet Conference 2021 (WIC 2021). I am very happy to be a core team member of GeaGraph.
- Our paper Heterogeneous graph neural architecture search got accepted by ICDM 2021, Auckland, New Zealand.
- Our paper "Adaptive optimizers with sparse group lasso for neural networks in CTR prediction" gets accepted by ECML-PKDD 2021.
- Won the championship on the Stanford University Open Graph Benchmark proteins dataset leaderboard: https://ogb.stanford.edu/docs/leader_nodeprop/#ogbn-proteins
- Our graph learning system paper won the Annual Best Paper Award (1st place) in Ant Group, 2021
- Our graph learning system was listed in 2020 Ant Technology Memorabilia (2020年蚂蚁技术大事记), 2021
- Our paper "Path-based Deep Network for Candidate Item Matching in Recommenders“ has got accepted by SIGIR 2021 as a full-paper.
- AI Smart Phone Recycling Robot (AI智能手机回收机器人)
- Our preliminary research on learning-to-compiling for deep learning is avaiable now: Woodpecker-DL: Accelerating Deep Neural Networks via Hardware-Aware Multifaceted Optimizations. 08/2020.