During this role, I managed the development of the Smart Tutor e-learning platform, a web application that allows users to upload text materials, create tests, conduct evaluations, and track performance based on defined metrics. As Project Manager, I led a team of 8, ensuring smooth team communication and workload distribution. I addressed challenges like framework selection, team disagreements, and deployment on Heroku. The platform was built using LangChain, OpenAI API, Django, PostgreSQL, Plotly, and Panel.In another project, I worked on deep learning models for predicting 16S rRNA gene copy numbers. The models included CNN, RNN, Transformer, and BERT, with the BERT model achieving an RMSE of 0.5, outperforming traditional bioinformatics tools. I handled challenges like imbalanced data and limited computing power by transitioning from PyTorch to TensorFlow and tuning models layer by layer.I also developed a graph neural network (GNN) model for predicting water depth levels in a simulated sewage network. I implemented four different GNN models, with the best one achieving an MAE of 0.0017. This project involved writing custom data loaders and developing graph neural network models using PyTorch Geometric.Technologies Used: LangChain, OpenAI API, Django, PostgreSQL, TensorFlow, PyTorch, PyTorch Geometric, Plotly, Panel

As a Visiting Scholar, I led and co-authored multiple research projects. One of the key projects explored the dissociation kinetics of small-molecule inhibitors in Escherichia coli. I discovered that the unbinding rate of Hoechst 33342 depends on the cell growth rate, with cell-to-cell heterogeneity leading to distinct subpopulations. To tackle challenges in measuring binding/unbinding rates at the single-cell level, I developed a new procedure using fluorescence microscopy. This research formed the basis of my first-author publication.Another project involved investigating active efflux systems in bacteria, which revealed heterogeneous bacterial responses to protonophores. This required advanced imaging techniques and collaborations with other research groups to overcome expertise gaps in fluorescence microscopy and motor speed measurement.Methodologies used: Single-cell microscopy, bacterial culture, molecular cloning, image analysis with Fiji and MicrobeJ, statistical data analysis and data visualization with Excel, Origin, SPSS, R

In my dissertation work, I explored the endocytic regulation of retinal pigment epithelial (RPE) cell polarity during eye development. I identified that the loss of Tsg101 severely disturbed RPE polarity, leading to irregular aggregates and altered cell adhesion protein distributions. This project involved complex tasks like managing genetically-engineered mouse colonies, performing immunohistochemistry, and collaborating with different facilities for advanced electron microscopy and electroretinogram (ERG) measurements.Methodologies Used: Immunohistochemistry, RPE Explant Culture and Staining, Cryosection, Electron Microscopy, ERG, Molecular Cloning, Statistical Analysis with Excel, SPSS, Image Analysis with Imaris Bitplane