Software Engineering Consultant
Singapore, Central Singapore, Sg
* Designed and implemented a Python PDF-to-Excel table extractor for use with 100+ bank statements, supporting table identification with no visibly separable lines, flattening of columns with stacked headers, layered subsection headers, proper assignment of rows split across pages, tables with overlapping columns, watermark removal, and other unconventional formats.* Leveraged Pandas, NumPy, SciPy, and scikit-learn libraries to optimize the PDF table extractor, improving total runtime by over 50% per page.* Designed and implemented an end-to-end deep learning pipeline for table extraction. Includes bank classification using a variety of CNN architectures, table bound and structure identification using Faster R-CNN and FCN-2s with OpenCV computer vision preprocessing, and table classification using NLP BERT multi-label classification models.* Wrote software to redact sensitive information from bank statements through direct editing of the PDF byte stream.* Implemented built-in table extraction verification mechanisms in which rows indicating total and subtotal values are compared against relevant rows to ensure that extraction was performed with 100% accuracy.* Integrated OCR for table extraction on scanned and read-protected PDF documents using custom-trained ABBYY and Tesseract models in a streamlined and automated manner.
