Engineered a hybrid bidirectional traceability system using Sentence Transformers and Pandas, implementing a dual-index architecture that maps unstructured PDF contract clauses to FOL representations via vector embeddings and cosine similarity. Enchanced transparency of computable contract data by developing a deterministic pipeline to classify modifications with explainable logic, validating the system's ability to align with original requirements through extensive testing on financial datasets.

Independently developed a machine learning algorithm using Scikit-learn and PyTorch to classify elastic stress contour images of barrier coating microstructures, assisted in postdoctoral research under RPI's department of Aerospace Engineering for optimal material selection. Employed convolutional neural networks (CNNs) to assess thermal images of materials within combustion chambers, resulting in a 250% workflow acceleration by replacing finite element simulations with ML-based predictions.