Faruk Pasalic

PDFMage is a Python library designed to extract text from PDF files. It offers a simple and efficient way to parse PDF documents and retrieve their textual content. The library supports extraction from specific pages of a PDF, providing flexibility for users. One of the key features of PDFMage is its configurable interface, which allows users to fine-tune the extraction process according to their needs. This includes setting parameters for the DBSCAN algorithm used in clustering words and columns.

Additionally, PDFMage includes debugging options that create visualizations of the extraction process, aiding in understanding and troubleshooting. The library also provides a Config class for customizing various aspects of the text extraction process, such as the path for output storage and the colors used in debug images. Overall, PDFMage is a comprehensive tool for PDF text extraction, offering high accuracy and customization options to cater to various user requirements.

I developed a configurable web scraping library. It is simple to use and provides several features for extracting data from websites. It supports different loading mechanisms (Python requests, Wb drivers). It can be configured to scrape a single article from the web, a list, or tables. It can also be configured to limit the number of elements to scrape and support scraping paginated data.

I worked as a senior software developer and architect on the project. My role was to discuss, develop, and maintain features for the data validation, forecasting, and reporting application. The system consists of two applications, one in Python, which is my main application. I also helped align features and minor code changes on another application.

I worked mostly as a consultant for a developing livestock scale. My main role was to investigate options for scales and RFIDs and to create the project's architecture. I investigated different hardware for solutions.

I developed an address extraction module involving data crawling, labeling, and preprocessing.

I also utilized CNN, RNN models, and multilayer LSTM on synthetic datasets from crawled websites and integrated features into the system.

I then implemented business name extraction using the Smith-Waterman algorithm, combining URL, title, and website content. I also created a website classification module using TensorFlow, with an LSTM-based classifier showing optimal performance among various models. Designing a profanity filter for the English language using the LSTM network with character-level and word-level embeddings demonstrated generalization even with distorted or misspelled words.

As a senior engineer, I led a Java and Spark project focused on ingestion of data from XML/JSON files into Spark, transformation of data for custom reports, and interfacing with external services.

My responsibilities included implementing Spark jobs, maintaining REST API, and documentation.

I led a small-scale project involving Python, Django, and React to develop a customizable calendar application.
My responsibilities encompassed creating custom functions, implementing a unique color scheme, and integrating other personalized features into the application.

I managed a web scraping project involving the extraction of textual and tabular data from websites. I then utilized PhantomJS, Selenium, and web drivers for data retrieval, processing, and persistence for subsequent analysis.