Reduced average time spent filling out an insurance application package from 12 minutes to 5 minutes (58% decrease) by reconstructing OCR data from Amazon Textract into interactable HTML pages.

Reduced average time spent per review from 80 seconds to 25 seconds (69% decrease) by developing a dynamic dashboard using React for clients to review accuracy of machine-learning models.

Developed and deployed video inpainting models from scratch for spatial and temporal gap-filling of large satellite image datasets which removed 99% of cloud covereage.

Researched efficacy of and implemented CNNs, GANs, RNNs, Transformers for deep learning with geospatial datasets.

Optimized data pipelines and parallelized training and testing loops across multiple GPUs which reduced training and testing times by 95%.

Developed backend of messaging service in Java with SpringBoot using MySQL database to hold user data, content. Used MyBatis framework to handle interactions with REST APIs.

Incorporated design patterns such as dependency inversion to prioritize adaptability, maintainability, and scal- ability by refactoring existing code.

Efficiently handle simultaneous CRUD calls by implementing caching and concurrency capabilities.