Bryan Wilkinson, PhD

Data Science


Natural Language Processing
Deep Learning for Natural Language Processing
Knowledge Representations


Ph.D. Computer Science, UMBC
B.S. Computer Science, UMBC



Recent Projects

Geospatial Intelligence Startup
Identifying and Forecasting Emerging Geopolitical Events on a Global Scale
Developed an efficient representation of news articles using a Universal Sentence Encoder-Transformer model, addressing processing requirements for performance, volume, and available resources. Created a ‘rare event’ forecasting model using Recurrent Neural Networks to identify geopolitical risks in localized areas. Devised a semantic frame extraction methodology to organize named entities, temporal expressions, and actors in documents.

Miner & Kasch
Data Science Training Courses
Created lectures, training classes, and hands on programming labs for training customers on a variety of subjects, including Spark, Machine Learning, Python, and Text and image analysis using Deep Learning,

Technical Expertise

  • NLP: Recurrent Neural Networks, Sequence-to-sequence Models, BERT, Word and sentence embeddings, Ontologies, Semantic and Syntactic analysis, Topic modeling
  • Convolutional Neural Networks
  • Machine Learning
  • Deep Learning Frameworks: Torch, PyTorch, Keras, Caffe
  • Python data stack: scikit-learn, pandas, numpy, Jupyter, others
  • Apache Spark and PySpark