Yuanyuan Zhang (张源源)
|
Director of Data Science and Data Platform Department,
Beijing Baixingkefu Network Technology Co., Ltd.(Series C unicorn company),
Beijing, China
E-mail: zhang.huanzhiyuan@gmail.com
|
About me
Yuanyuan Zhang received a B.S. degree from Jilin University in 2012 and has done pioneering work in the
industry in many leading companies in the track, such as Baidu, Ledongli, Alibaba, and People's Car Service.
He has accumulated more than 10 years of work in data science and data platforms in mobility data, trajectory
data, LLM
and other fields.
During my stay at Baidu, I was one of the annual best team members at Baidu Company..
During my stay at Ledongli, I was independently responsible for the data science work of the first annual
selected app in the Apple Appstore in China.
During my stay at Alibaba, the AI sports product was the first application in the industry to count fitness
movements in real-time on a mobile phone. During the epidemic, it helped hundreds of thousands of college
teachers and students successfully carry out physical education teaching and promoted it to tens of millions
of primary and secondary students; Responsible for the commercial algorithm work of walking and running, with
an annual income of nearly 100 million yuan.
During the People's Car Service period, I led the team in developing an industry-leading SDK and UBI solutions
for dangerous driving behavior identification. We applied for 7 patents and published 10 top-tier conference
papers, including 9 CCF-A papers. This project became the first in the insurance industry to be selected for
the supervision sandbox of the People's Bank of China for financial technology innovation applications. I
also led the team in developing the industry's first scalable, high-performance, cloud-native, one-stop
geospatial data platform, aiming to address the
issue of
missing data/AI infrastructure in the field of geospatial data, this scheme has also been selected by Apache
iceberg as the official scheme supported by geometry data. Currently, I am leading the team in developing
a full-process agent system for vehicle insurance based on Large Language Models (LLMs).
I am deeply passionate about language models (LMs) that are designed to meet a broad spectrum of human needs
and center around human-centric concerns. My focus revolves around several critical questions:
-
How can we develop robust and scalable methods to evaluate LMs, particularly for tasks where human
performance is challenging?
-
What approaches can we employ to create LMs that encapsulate the full spectrum of human desires,
including ethical values,
personal preferences, and specific competencies?
-
Is it possible to tackle complex issues by enhancing the interactive capabilities of LMs? This enhancement
could include
augmenting tools and memory, facilitating multi-agent communication, and fostering effective human-AI
collaboration.
I am convinced that these areas of evaluation, alignment, and interaction are interconnected, forming a
synergistic loop
where each element influences and bolsters the others. My aspiration is to delve into each of these realms not
only as
isolated topics but also in terms of their interplay, with the ultimate goal of crafting general-purpose
language models that are truly human-centric.
Research
My research interests include:
-
Language models (LMs)
-
Computational Statistics
-
Reinforcement learning
Publications
-
Zhang, Y.; Zhao, K.; Chen, Z.; Zhang, Y.; Du, Y.; and Lu, X. "A Graph-based Representation Framework for
Trajectory Recovery via Spatiotemporal Interval-Informed Seq2Seq." Accepted by IJCAI 2024.
-
Zhang, Y.; Du, Y.; Zhang, Y.; and Zhang, Q., Moral Hazard and Transparency in
Peer-to-Peer Auto Insurance with Telematics
, ICIS 2023 Proceedings. 19
-
Zehong Zeng; Yueyang Liu; Xiaoshi Lu; Yuanyuan Zhang; Xiaoling Lu, An Ensemble Framework Based on Fine Multi-Window
Feature Engineering and Overfitting Prevention for Transportation Mode Recognition
, UbiComp '23
-
Shiyao Huang; Junliang Lyu; Sinian Zhang; Ruiying Tang; Huan Xiao; Yuanyuan Zhang; Xiaoling Lu, A Post-processing Machine Learning for Activity
Recognition Challenge with OpenStreetMap Data
, UbiComp '23
-
Jiebi Deng; Jingqiu Xu; Zicheng Sun; Danning Li; Hongxuan Guo; Yuanyuan Zhang; Xiaoling Lu, Enhancing Locomotion Recognition with Specialized
Features and Map Information via XGBoost
, UbiComp '23
-
Mengyuan Li; Jun Zhu; Yuanyuan Zhang; Xiaoling Lu, Enhanced SHL Recognition Using Machine Learning
and Deep Learning Models with Multi-source Data
, UbiComp '23
-
Yaya Zhao; Lin Song; Cheng Ni; Yuanyuan Zhang; Xiaoling Lu, Road Network Enhanced Transportation Mode
Recognition with an Ensemble Machine Learning Model
, UbiComp '23
-
Hanchao Yan; Xinran Huang; Yiling Ma; Ruizhe Yao; Zhiyu Zhu; Yuanyuan Zhang; Xiaoling Lu, AttenDenseNet for the Sussex-Huawei
Locomotion-Transportation (SHL) Recognition Challenge
, UbiComp '23
-
J Su, Y Zhang, "Triple-O for SHL Recognition Challenge: An Ensemble Framework for Multi-class Imbalance
and Training-testing Distribution Inconsistency by OvO Binarization with Confidence Weight of One-class
Classification", UbiComp '21, September 2021, Pages 401–407 [pdf]
-
Y Duan, Y Zhang, C Gao, M Tong, Y Zhang, K Bian, W Yan, "Trajectory-matching prediction for friend
recommendation in anonymous social networks", GLOBECOM 2017-2017 IEEE Global Communications Conference,
1-6 [pdf]
Campus Experience
Master Tutor (Industry), Department of Applied Statistics, School of Statistics, Renmin University of China, 2020~2025
-
Give undergraduate students the course "Data Science Practice" and teach the part of graph neural
network, which was highly praised.
-
Tutor many students in scientific research activities, and has published 9 CCF-A papers.
-
Guided students in participating in the NeurIPS
Large Language Model Efficiency Challenge, achieving a top ten placement.
Bachelor, Information and Computing Science, Jilin University,
06.2012
-
In the first three academic years, the professional course score is in the top 10% of the major
-
Main Courses: Calculus, Linear algebra, Probability and Statistics, Real Analysis, Numerical
analysis, Partial Differential Equations,
Information
Theory, Data Structures and Algorithms.
Competitions and awards
-
National Encouragement Scholarship, 2010~2011
-
First-Class Scholarship, Jilin University, 2009~2010
-
National Encouragement Scholarship, 2008~2009
-
Meritorious Winner in the Mathematical Contest In Modeling, 02.2010
-
Top 10 in Computational Mathematics in Peking University's Direct Doctoral Mathematics Examination,
05.2011
-
Both teams I coached ranked in the top 10 at the Sussex-Huawei Locomotion
Challenge 2021,
10.2021
-
All six teams I mentored qualified for the Sussex-Huawei Locomotion
Challenge 2023 finals, with four placing in the top 10,
10.2021
-
Top 10 in NeurIPS Large Language Model Efficiency
Challenge,
12.2023
Work Experience
-
Director of Data Science and Data Platform Department, Beijing Baixingkefu Network Technology Co., Ltd.,
09.2020-Present
-
The industry-leading dangerous driving behavior recognition system keeps ahead of competitors in many
key links such as event recognition, scene recognition, driver and passenger identification
-
The industry's first scalable, high-performance, cloud-native, one-stop geospatial data platform
-
Developing a full-process agent system for vehicle insurance based on Large Language Models (LLMs).
-
Staff Engineer, Alibaba, 09.2017-09.2020
-
Commercialization of mass sports data such as walking and running
-
The electronic coupon allocation based on purchase intention estimation and MCKP, which also inspired a
work of 2020 CIKM
-
The industry's first mobile fitness action real-time counting and timing system
-
Data scientist, Ledongli Co. LTD, 05.2014-09.2017
-
The industry's first mobile pedometer that can run continuously in the background using IMU and
Magnetometer
-
Automated human activity recognition system based on motion sensor
-
Mobile automatic wake-up system with power saving and timely
-
Forecast of sales lead conversion possibility
-
Internship software engineer/Software Engineer, Baidu, 08.2011-05.2014
-
Web Traffic Forecasting using ARIMA
-
Session-aware document recommendation
-
Session-aware Personalized music recommendation
Social Experience
|