My portfolio for Machine Learning and Data Science.

• Modified CycleGANs to enchance the OCR on Electronic Health Records (EHRs).
• The data was utilized from KEPRO where I was interning for the Summer of 2020, the data is not accessible for the general public due to HIPAA and other privacy protection laws for EHRs.
• The models where trained on Azure Machine Learning Services, the denoised documents were passed through Tesseract which is an open sourced OCR software.
• The advantage of doing OCR on the EHRs is to create an unstructured dataset to do complicated NLP downstream tasks like such Summarization, NER and QnA using Bio Bert/ Clinical Bert.