Projects

1. Coast Guard Harbor

   • Build infrastructure-as-code modules to configure Databricks resources resulting in a 99.95% reduction in the on-boarding time for new users

   • Develop terraform tooling to manage data privacy, role based access controls, data sharing agreements

   • Build out proof of concept machine learning experiments using spark and PyTorch demonstrating ways to save costs across the coast guard

2.  Coast Guard HR Migration

   • Migrate coast guard HR data and reporting from legacy databases to databricks, reducing the time of report development from 6 weeks to 24 hours

   • Recognized with an internal award for highest quality deliverables 

3. Steampunk AI Lab

   • Develop prototype using DSPy library demonstrating the use of small models for mathematical reasoning

   • Critical contributions during a tech challenge resulting in a $47M multi-year award

Projects

1. Submarine data analytics platform 

   • Develop error logging framework to streamline errors from multiple sub-systems to one single point of entry reducing debugging time by 50%

2. Machine learning reporting tool

   • Refactor researchers' experiments to production predictive machine learning software 

Projects

1. RAPID Data Platform

   • Built a data engineering product from design to production using Snowflake, Apache NiFI and AWS earning the company a record $15M contract award

     1. 1. Ingested over 500 gigabytes of data daily from 15 different data vendors

   • Optimized thousands of stored procedures in Snowflake to reduce cloud expenditure

   • Interfaced with the data science team to bring into production tools for anomaly and fraud detection

   • Hired a team of developers as the product matured enabling us to rapidly develop new features resulting in being recognized with an award for best teamwork

2. Trafficking evidence search

   • Built a containerized search engine to rapidly sort through troves of financial evidence linked to human trafficking chains using Docker reducing analysis time by 30 hours per week

Projects

1. Built a support-vector-machine classifier to predict construction codes and materials aiding Hurricane Maria recovery efforts achieving an 82% precision rate

2. Maintained an enterprise HDFS cluster managing upgrades to services like Hive and Spark for over 250 researchers working on critical national security missions

3. Parallelized an application using AWS EC2 instances to reduce runtime by 99.96% for an application for regular reporting to regulatory agencies