Yufeng Yang Email & Phone Number

United States

Palo Alto, California, United States

Compiled benchmark datasets of satellite imagery for pretraining and fine-tuning; Pipelined the process of image-text generation using LLM API; Improved the quality of image-text pairs by incorporating geospatial analysis and different tokenization approaches; Dived deep into frameworks of vision transformer backbones (Vit/Swin/DINO) and VLMs.

Twin Cities, Minnesota

Conducted systematic literature review on socio-environmental perforamnces of two US-promoted climate-smart agricultural practices. Developed meta-random-forest models to generate spatially-explicit predictions of carbon offsets and economic outcomes. Mapped optimized pathways based on cost-benefit analysis (in US$) to inform policy making.

Fellowship research at UM Informatics Institute. Pretrained a baseline ML model (GRU-SOC) to simulate SOC change dynamics and crop yield using large quantity of synthetic data. Designed knowledge-guided ML models (KGML-SOC) that outperformed the baseline by introducing a hierarchical structure, a customized loss function, and intermediate variables.

Calibrated and validated process-based model performances on soil water and temperature, and CO2 and N2O fluxes. Identified spatial-temporal drivers and variations of N2O emissions in the US Midwest and and drivers using causal inference (PCMCI). Predicted carbon mitigations and social benefits under 12 climatic × management scenarios.

Derived time-series of vegetation indices (NDVI, CGVI, EVI) using Sentinel-2 satellite imagery. Obtained the peak GCVI as an indicator for oak yield with using harmonic regression. Delineated field management zones of high, medium and low production potentials using k-mean classification.

Doctor Of Philosophy - Phd, Agricultural Engineering, Gpa 3.95/4.0

Bachelor'S Degree, Hydraulic Engineering

Master'S Degree, Hydraulic Engineering