Graduate Research Assistant, Tufts University – Aldridge Lab (multi-stage regression, multimodal VAE, microscopy-to-morphology pipeline)
Built machine-learning pipelines that perform image analysis and derive morphological outputs from microscopy-derived data, enabling downstream labeling/annotation workflows from raw imagery. Developed and calibrated predictive models (Random Forest/XGBoost) and multimodal models (morphological profiling + RNA-seq) to generate treatment-mechanism and in-vivo performance signals used as training/evaluation targets. Deployed these models and pipelines on HPC infrastructure and in a user-facing web application to support ongoing dataset processing and model evaluation. • Implemented an end-to-end microscopy-to-morphology processing pipeline (Python, Bash, SLURM, neural network inference) on large microscopy datasets. • Trained Random Forest/XGBoost regressors to predict in-vivo performance from in-vitro features. • Developed and trained a multimodal variational autoencoder for mechanism-of-action discovery. • Built a Flask/SQL-backed web app to deliver image analysis results to users.