My research focuses on developing machine learning-powered microfluidic diagnostics to rapidly detect carbapenemase-producing organisms (CPOs). By integrating colorimetric detection with ML algorithms, we achieve high-accuracy classification of antibiotic-resistant bacteria in minutes rather than days. Using recursive cross-validation and stratified k-fold methods, our optimized k-nearest neighbors (KNN) model reaches 100% accuracy for certain carbapenemase classes within just 5 minutes.

We are currently implementing deep learning for microfluidic image analysis, employing convolutional neural networks (CNNs) with transfer learning to automate classification from smartphone-captured images. This work is paving the way for faster, more accessible clinical diagnostics to combat antibiotic resistance in healthcare settings.

During my undergraduate research, I worked on enhancing the detection sensitivity of Francisella tularensis, the causative agent of tularemia. This project involved pre-concentrating F. tularensis lipopolysaccharide (LPS) using Invitrogen Dynabead M-270 Epoxy magnetic particles before analysis with lateral flow immunoassay (LFI). This approach aimed to improve diagnostic capabilities by increasing sensitivity with larger sample volumes. The significance of this work lies in developing an effective, affordable point-of-care diagnostic tool for tularemia, which is particularly valuable in outbreak scenarios or resource-limited settings.