πŸ“š Education

Boston University β†—

πŸ—“ 09/2022-Present πŸ“ Boston, MA

  • Pursuing Master of Science in Computer Science, GPA 3.96/4.0

South China University of Technology β†—

πŸ—“ 09/2017-06/2021 πŸ“ Guangzhou, China

  • Bachelor of Engineering in Software Engineering

🧰 Work & Internship

Intel Asia-Pacific Research & Development Ltd (Full Time) β†—

πŸ—“ 08/2021-08/2022 πŸ“ Shanghai, China

  • AI Framework Engineer in the Artificial Intelligence and Analytics Department, SATG

  • Contributed to the development of oneDNN’s backend Graph Compiler, a mathematics acceleration library

  • Enhanced deep learning speed through fusion, loop rescheduling, graph inline, and optimization for CPUs supporting the AVX512 and AVX512-BF16 instruction set

  • Implemented element-wise operators such as activation functions, as well as graph operators including normalization and VNNI reordering. Created graph patterns like MHA to analyze performance bottlenecks and optimize

Tencent - WeChat Group (Internship) β†—

πŸ—“ 06/2020-09/2020 πŸ“ Guangzhou, China

  • Software Engineer in WeCom team

  • Full stack developed an internal-use test case management system with Python Tornado and React

  • Created internal data visualization panels using Grafana

πŸ”¬ Project & Research

Accelerating Large-Scale GNN Training with Programmable SSDs β†—

πŸ—“ 09/2023-Present πŸ“ Boston, MA

  • MS thesis under the supervision of Professor Vasiliki, focusing on FPGA programming using Vitis HLS for direct data sampling from SSDs.

  • Bringing computation closer to SSD, conserving PCIe bandwidth for large-scale GNN sampling.

  • Anticipate achieving higher performance compared to current CPU sampling methods, with potential for linear scalability.

Advanced Matrix Multiplication - Strassen Algorithm β†—

πŸ—“ 02/2023-05/2023 πŸ“ Boston, MA

  • Implemented and optimized the Strassen algorithm using OpenMP and SIMD intrinsic

  • Conducted performance analysis and comparison of different matrix multiplication implementations

Fine-Tuning Pretrained Vision Transformer Models on the iNaturalist Dataset β†—

πŸ—“ 02/2023-05/2023 πŸ“ Boston, MA

  • Fine-tuned models based on NAT, DiNAT, MaxViT, and DaViT, achieving image classification accuracy comparable to state-of-the-art models

  • Analyzed structural differences among various vision transformers

  • Compared the outcomes of fine-tuning with those of training from scratch

Online Signature Verification Based on Dynamic and Static Features β†—

πŸ—“ 01/2021-06/2021 πŸ“ Guangzhou, China

  • Extracted static features from signature images using a CNN-based Autoencoder network and dynamic features from signature trajectory using a GRU-based Autoencoder network

  • Fused features using early and late fusion techniques

  • Employed a CNN-based Siamese network for similarity judgment to determine signature authenticity

Parallel Programming of “Group-By” Operation β†—

πŸ—“ 03/2020-06/2020 πŸ“ Guangzhou, China

  • Simulated the “Group-By” operation of databases and calculated the group number of random numbers

  • Grouped data based on prefix and used OpenMP for parallel processing within the same computing node

  • Used MPI for cross-node parallel operation, distributing the computation task to each node and aggregating the results - this improved efficiency several times compared to serial programs

πŸ—“ 09/2019-11/2019 πŸ“ Guangzhou, China

  • Designed an approximate detection method to search for short video clips in long videos

  • Used SIFT to extract feature points and FAISS vector database to build a long video feature library

  • Performed accurate matching using hierarchical matching and sliding window, achieving an accuracy within a 5-second range

UWP-Based Property Management System β†—

πŸ—“ 06/2019-09/2019 πŸ“ Guangzhou, China

  • Created frontend of property management system using C++ and Universal Windows Platform.

  • Developed backend with Python and Django, implementing RESTful APIs for communication.

πŸ† Honors & Awards

1st Scholarship of South China University of Technology β†—

πŸ—“ 10/2020

  • Awarded to meritorious students based on academic excellence

2nd Prize of CCF National Olympiad in Informatics in Provinces β†—

πŸ—“ 11/2015

  • Algorithm competition akin to ACM-ICPC

πŸ“± Contact Information

  • Email: i @ syh . one

  • GitHub：https://github.com/Souukou