First authored a top-tier journal paper under review at Nature Machine IntelligenceImproved predictions in the extreme few shot learning setting on sim-to-real fusion experiment dataDesigned a novel Vision Transformer based Masked Autoencoder to handle multi-modal dataIncreased prediction speed by over 4000x by using deep surrogate models for scientific simulationsBuilt a Chatbot (Llama2) with Retrieval Augmented Generation using Kubernetes, Docker, and FastAPIUsed ChatGPT4-V to tune visualizations parameters (e.g., UMAP/T-SNE) with no human intervention

First authored a top-tier journal paper in AI on explaining RL agents using deep generative modelsFirst authored a top-tier conference paper in VIS to identify/explain dataset shift to non-expertsCo-authored four conference papers, helping code deep vision models in PyTorch and plan user studiesDesigned user studies and developed user interfaces using C#, JavaScript, D3, and FlaskCreated a state of the art loss function for improving open set detection in neural networksReviewed dozens of conference/journal paper submissions, top reviewer for NeurIPS 2022Founded and led the Artificial Intelligence Graduate Student Associations with over 200 members

First authored a paper for CVPR on understanding and auditing GANs beyond summary statisticsDesigned new algorithms for identifying unique and shared attributes between two datasetsBuilt state of the art generative models for approximating data distributions (e.g., StyleGAN2)

Improved semi-supervised multi-label accuracy of Saskatchewan court cases by 18% using LongFormerReduced error of part manufacturing time predictions by 37% using tabular data with Pycaret

Won the robotics competition at the Unix 50th anniversary international eventUtilized state of the art language models to perform 4G/5G patent classificationDeveloped new deep learning models to process arbitrary length sequential data

Developed and maintained .Net desktop application for a multi-million dollar fleet of printer test tools. Became team lead for the motion control software on the R&D and production tools. Collaborated with different teams in India, Singapore, Barcelona, and Ireland to develop new Thermal Inkjet technologies. Travelled to Singapore to perform tool installation and training. Created Software for real-time, rotary encoder, error correction.

Co-authored a paper accepted at IEEE International Workshop on Debugging and Repair 2018. Investigated Delta Debugging for the Siemens suite using Python and bash. Calculated Software Fault Localizations coefficients for the suite using Java