Conner Adams Email and Phone Number

• Trained, validated, and deployed convolutional neural networks for medical image characterization. Models trained in Tensorflow on a GPU enabled EC2 server on Amazon Web Services (AWS). Deployed on the edge in Surgimap software, and via an API online. These networks improved user interaction with our software by automating steps involved in surgical planning, driving more surgeons to use our platform. Also, once deployed across our entire image database they allowed our team to quickly query those data to access images they would previously need to manually search for.• Productionalized ETL applications using PySpark and Airflow on AWS. Transformed ~5 terabytes of unstructured image and text data to characterized data to be used in model development.• Developed feature-engineering regimes and unsupervised clustering models for unstructured medical record text data. Used vectorization as well as neural network based text embedding. Models trained to group textual observations to patients achieved ~99% accuracy. These tools were then used in large scale data extractions, eliminating the need for labor and resource intensive manual data cleaning.• Designed interactive dashboards for various stakeholders using AWS Quicksight, allowing a clear view of key parameters of our machine learning training datasets, data pipelines, and user engagement.

• Analysis of large volume data from National Centers for Environmental Prediction (NCEP) re-analysis.

• Developed numerical analysis package for estimating cortical current source density.• Wrote and implemented Markov Chain Monte-Carlo as well as Multi-Start Gradient Descent to determine noise-optimal parameters of our model (MATLAB).