Kenneth D Email & Phone Number

Ashburn, Virginia, United States

- Engineered and deployed document parsing and question-answering systems, harnessing advanced natural language processing techniques. Achieved precision in extracting pertinent information from unstructured data sources using the Vision-based Encoder-Decoder Model.- Contributed the design and execution of data pipeline utilizing Python libraries such as Pandas, Numpy, and Spark. Ensured seamless data integration across diverse platforms, contributing to enhanced data accessibility and system efficiency.- Leveraged Pandas, Seaborn, and Tableau to craft engaging visualizations. Transformed complex datasets into intuitive visual representations, aiding in effective communication of insights to both technical and non-technical stakeholders.- Led the development and implementation of data pipeline jobs using Spark, leveraging its robust capabilities for large-scale data processing. Designed and optimized complex ETL (Extract, Transform, Load) workflows to ingest, clean, and transform data from various sources, ensuring data integrity and consistency. Utilized Spark's distributed computing framework to efficiently handle massive datasets, significantly improving processing speed and scalability. Collaborated with cross-functional teams to deploy and monitor data pipelines, providing actionable insights for business decision-making.

New York, New York, United States

- Improved extraction accuracy significantly by strategically integrating multiple Graph Convolutional Network (GCN)-based Knowledge Information Extraction (KIE) models, specifically incorporating BERT, into an advanced transformer-based Large Language Model (LLM) architecture using PyTorch and the LangChain library.- Proficient in Named Entity Recognition (NER) and summarization techniques, with a proven track record of implementing them in Intelligent Document Processing (IDP) projects- Led the development of a new Convolutional Neural Network (CNN)-based Text Detection model (in OCR process), seamlessly integrated into our NLP system. This innovation led to a remarkable 41% reduction in inference time for extracting critical information from scanned documents and card images.- Using Jenkins/Airflow to manage data pipeline, efficiently scheduled web scraping tasks, data extraction, and indexing operations, resulting in a reliable and up-to-date knowledge base for our users.- Used Git for version control, including repository management, collaborative development, conflict resolution, and implementing branching strategies, ensuring organized and efficient software development processes.- Collaborated closely with cross-functional teams, including data scientists, software engineers, and DevOps specialists, to create a robust and scalable MLOps pipeline.

New York, New York, United States

- Created an innovative solution for NLP project, leveraging denoising, text detection, and tesseract OCR to detect textual content and tabular structures in scanned certificate documents and PDFs. This groundbreaking approach led to an impressive 54% reduction in inference time, revolutionizing document processing efficiency.- I designed an economic simulator to analyze time-series data, including factors such as inflation rates, employment rates, GDP growth, and trade balance. This involved employing various methods, such as Prophet, ARIMA, LSTM, and others, to gain insights and make informed economic predictions.- Proficiently utilized both PyTorch, Pandas and PySpark as integral components in orchestrating distributed computing processes, ensuring seamless data handling (JSON) and computation across large-scale datasets.- Incorporating Docker containerization, Kubernetes, RCP(Runtime compute platform), and Databricks into our development workflow streamlined application deployment and ensured consistent environments across development, testing, and production, simplifying the entire release process.- Led the implementation of advanced denoising and deraining algorithms, strategically transitioning from VGG-16 to MobileNetV3 to optimize processing time (CUDA) while maintaining top-tier performance in image enhancement tasks.- Conducted comprehensive benchmarking of cutting-edge logo detection models (YOLO, SSD, ResNet, FastRCNN), meticulously evaluating their performance within the TensorFlow framework.- Leveraged AWS GCP, and DataBricks to establish and oversee the essential infrastructure for a logo detection project, ensuring a streamlined and automated deployment process.- Demonstrated proficiency in C++, OpenCV, and advanced data augmentation techniques to generate synthetic training data for logo detection