Developed a CNN and RNN based classifier to predict the topic of a conversation to provide better recommendation in the Voice CoBot product.Engineered the fine tuning of BERT to remove the dependency of regex classifiers from the system.Maximized annotation data count from raw data with the usage of techniques like semi-supervised learning and active learning.

Provided Machine Learning/Computer Vision assistance for a team working with NASA and SpaceX to send microbiological experiments (NemaFlex) to the International Space Station with an upcoming launch.Developing and deploying a script for detecting C. Elegans inside the NemaFlex, using efficient Computer Vision techniques in MATLAB and auto-crop the experimental videos based on the nematode’s movement.Implementing a python script to detect unexpected spurious peaks and depressions in the data from the NemaFlex experiment by employing efficient outlier detection algorithms.Created a script that converts tiff stack from a path to AVI file and was modified to handle exceptions to enable batch processing.

Collected data from different sources within the organization and created a dataset.Designed and developed grading and classification medical images using deep learning architectures such as VGG16 and Inception with 90% accuracy, sensitivity, and specificity.Compared the above model with other algorithms such as ResNet, InceptionResNetV2, Xception and VGG19.Separated set of maps from a series of layered images for semantic segmentation algorithms.Developed a deep learning algorithm(CUDA) using U-Net architecture for medical image segmentation with 88% accuracy.Wrote various annotation manuals for different products in the organization.

Grader for Programming Principles I and II (lab and class).Determine grades based on personal judgment and knowledge of the subject.Automated the grading task by developing python scripts which assisted professors and reduced the workload by 40%.

Responsible for continuous development of the web applications in an Agile environment, based on the requirements from the department by writing short procedures in SQL server and increase the efficiency by 65%.