As the lead developer and full-stack engineer, the PhenoSight project was spearheaded to achieve end-to-end software development. This comprehensive image analysis tool was created to leverage machine learning and image processing algorithms for the analysis and visualization of cellular phenotypes linked to specific genetic perturbations. Implementing a diverse range of machine learning techniques such as SVM, logistic regression, decision tree, and random forest classifiers, the goal was to accurately classify mock cells and perturbed/infected cells. Rigorous evaluations were performed on each classifier, utilizing F1 scores and ROC-AUC curves to ensure optimal performance throughout the development process.Tool highlights:- Perform interactive field-of-view (FOV) - level annotation- Visualize and fine-tune automated segmentation of cellular objects using AICS and cellpose segmenters- Interactively analyze and visualize top perturbed cell features e.g. area, intensity, texture

Relational Learning for DNA Sequencing: Developed a base-calling framework based on statistical relational learning to sequence DNA and RNA strands from MinION nanopore current signals after appropriate pre-processing of signals. This method increases accuracy by 5% over the existing models in sequencing longer reads.Nanopore Base-caller: Improved the signal-to-noise ratio in solid-state nanopore devices via the enhancement of the in-plane nucleotide currents by translocating synthetically-modified nucleotides for DNA/RNA sequencing. The current signal data is statistically analyzed, processed, and fed into a supervised machine learning framework for accurate base-calling.DNA Data Storage: Developed a complex computational biological model to detect and map breaks in the backbone of DNA strands using MoS2 nanopores for DNA data storage. This project resulted in two first-authored publications and patent.Classification of Cancer Biomarkers: Developed an efficient model which combines signal processing and machine learning algorithms with atomistic molecular dynamics code and quantum transport calculations for detection and classification of epigenetic biomarkers using ultra-thin nanopores. This project resulted in a first-author publication and a patent.Massively Parallel 2D Nanopores for DNA Detection: Developed computational tools for simultaneous detection of multiple DNA strands translocated through massive array of electrically active 2D solid-state nanopores. This technique overcomes the shortcomings of existing technologies paving way for the realization of next generation nanopore devices.

Collaborated with the Computational Nanotechnology Group to develop software for simulation of single molecule detection techniques using two-dimensional solid-state nanopores.