Graduate Research Assistant - Cornell University - Ithaca, NY
(2021-08)
PhD Projects
- Co-developed a linear mixed-model model predicting cross performance from additive and dominance effects; outperformed standard breeding-value selection across 80 simulated scenarios over 40 cycles.
- Validated the pipeline on three years of yam field data (2,302 progeny); selected clones outperformed unselected material on all four traits (p < 2.2e-16).
- Applied the tool in an operational breeding program, selecting 158 clones at 6.9% intensity and shortening the breeding cycle to three years.
- Conducted research on the effectiveness of smoothing splines in accounting for spatial variation in agricultural field trial analyses; analyses showed improvements in genetic parameter estimation by up to 13%
Graduate Research Assistant - Cornell University - Ithaca, NY
(2025-01 - 2025-05)
Introduction to Computer Vision
- Utilized Contrastive Language-Image Pretraining(CLIP) neural network model to classify images
Graduate Research Assistant - Cornell University - Ithaca, NY
(2024-09 - 2024-12)
Data Mining and Machine Learning
- Collaborated with a team of 4 to predict customer churn using Tree-based machine learning models, especially XG-Boost in Python
- Implemented Feature Engineering, Exploratory Data Analysis, and Hyperparameter tuning
Graduate Research Assistant - Cornell University - Ithaca, NY
(2022-01 - 2022-05)
Quantitative Genomics and Genetics
- Utilized and compared Bayesian and ridge regression models for genomic prediction on a synthetic population
Graduate Research Assistant - CIMMYT - Nairobi, Kenya
(2019-01 - 2020-07)
MSc Projects
- Analysed 380–1,400 genotypes against up to 278,810 predictors, fitting mixed models to separate genetic signal from environmental and trial effects.
- Used principal component analysis with kinship matrices to correct for population structure; three components captured up to 72% of genetic variance.
- Detected 103 significant marker-trait associations in total, each explaining 3–12% of phenotypic variance for the traits studied.
- Built ridge-regression and covariance-shrinkage prediction models reaching 0.72–0.78 out-of-sample accuracy under repeated k-fold cross-validation.
