Yu Cheng 鄭嵎

Yu Cheng 鄭嵎

Sustainability Data Scientist

MIT Office of Sustainability

Biography

Yu has 10 years of experience using pioneering high-resolution climate models to study the ocean and its role in the climate system. Coming from Taiwan, he is a big fan of all weather phenomena and the latest technology. With an atmospheric science background, he is passionate about combining his experiences in climate modeling, software engineering and data science to solve more challenging problems.

Fun facts: Yu loves all sports with racquets. He plays in local tennis leagues school badminton clubs and has been stringing racquets for friends for many years. One of his dearest memories from Miami was serving as a ballperson at Miami Open and handing towels to Rafael Nadal.

Interests
  • Data-driven Organization
  • ML/AI Application
  • Predictive Modeling
  • Data Visualization
Education
  • PhD in Meteorology and Physical Oceanography, 2018

    University of Miami

  • BSc in Atmospheric Sciences, 2010

    National Taiwan University

Experience

 
 
 
 
 
Sustainability Data Scientist
September 2023 – Present Cambridge
Innovate MIT’s sustainability data collection, reporting, and sharing efforts, leveraging cloud technology and modern data stacks.
 
 
 
 
 
Founding Senior Data Scientist
Web3 Builders
June 2022 – April 2023 Boston
Integrated disparate data sources to develop labeled datasets for ML models, predicting scam probabilities of Ethereum smart contracts and wallets. Established ETL and ML pipelines using cloud infrastructures.
 
 
 
 
 
Senior Data Scientist (Consultant)
October 2021 – June 2022 Boston
  • Implemented and deployed ML solutions tailored to clients’ objectives, resulting in improved business performance through meticulous exploratory analysis, metric definition, and continuous evaluation of data science life-cycle outcomes.
  • Orchestrated a CNN-based illegal fishing vessels detection and labeling system, designing and implementing end-to-end data workflows for large-scale data ingestion, processing, tagging, and publishing.
 
 
 
 
 
Senior Atmospheric Data Scientist
October 2018 – October 2021 Boston
Created novel weather products by integrating computer vision and ML techniques, resulting in a near real-time global precipitation pipeline using geostationary satellite images and cloud-based technologies using Google Cloud Platform, xarray, dask and satpy.
 
 
 
 
 
Research Assistant
RSMAS, University of Miami
July 2012 – June 2018 Miami
  • Explored Agulhas Leakage variability through a state-of-the-art high-resolution climate model using Lagrangian particle analysis and open-source tools, resulting in 4 peer-review articles.
  • Organized and led weekly discussion sessions, developed the final exam problems for the course Current topics of Weather and Climate, graded weekly quizzes, exams, and final term-papers.
  • Coordinated with the instructor to synthesize news articles regarding climate change, mitigation, and policy to be covered in the lectures for the course Climate and Global Change

Recent Posts

Projects

Climate changes in the high-resolution 20th century simulation

Climate changes in the high-resolution 20th century simulation

We noticed that Agulhas leakage is indeed higher in the year 2000 CO2 level control runs, but no significant increasing trend in the climate change simulation. It might be due to the equilibrium timescale of westerlies to changing CO2, or an internal bias of our high-res CCSM.

Large-scale forcing dominates interannual variability of Agulhas leakage

Large-scale forcing dominates interannual variability of Agulhas leakage

The interannual variability of Agulhas leakage is accessed in an ocean eddy resolving coupled simulation to test the hypothesis that, on such timescales large-scale forcing dominates leakage variability, regardless of eddy structures.

Quantifying Agulhas leakage in a high-resolution coupled climate model

Quantifying Agulhas leakage in a high-resolution coupled climate model

The senstivity of leakage estimates to the model output velocity frequency needs to be addressed before we can apply it to investigate its link to climate variability.

Publications