Optimization and training novelty:- Conducted state-of-the-art research on degraded historical document analysis.- Developed Deep Learning-based algorithms to reconstruct degraded documents from images.- Proposed new Deep Learning training procedures to be applied to document analysis.- Proposed a self-supervised approach for learning from limited data. - Formulated the problem of restoring historical documents as a multi-objective optimization problem. - Published multiple research papers which included a paper that was accepted at ICPR 2022.Data collection:- Gathered and cleaned data from different sources.- Developed a simple procedure to synthesize data which helped boost the training accuracy. Applications:- An end-to-end reconstruction tool that can take degraded documents and output a readable version.