• Developed enterprise solution to match generically structured entities with incomplete queries

• Incorporated semantic search capability with word embeddings dropping failure rate by 40%

• Parallelized ML training workflows dropping vectorization time from three hours to 20 minutes

• Designed training data persistence system to accept, curate, and store feedback in SingleStore DB

• Embedded engine into broader database update workflow, alerting users to potential database duplicates

• Built Prometheus based monitoring system to validate engine uptime without false flags

• Developed solution to identify matching pairs of legal entities from separate entity datasets

• Implemented linguistic metrics (such as Levenshtein Dist.) to quantify the similarity between entity traits

• Used machine learning to form match predictions from metrics scores for over 95% accuracy when tested on over 5,000 ground truth data points

• Led production of web interface to search for legal owners of clients to expand business operations

• Developed algorithms to consolidate and sort potential ownership matches for succinct UI responses

• Interfaced with internal business users to ensure the application led to new client relationships for GS

• Interfaced Velodyne LiDAR and camera data with ROS in Nvidia Jetsons for an autonomous vehicle data logger (Autonomous ONE)

• Added ROS functionality to automotive data logger software (VSpyX)

• Developed programs to utilize built in Jetson encoders to record, store, and stream video over Ethernet