Dragan Stosic

At Rival (ex-SilkRoad Technology), I built and shipped the AI platform behind enterprise HR automation for Netflix, eBay, Cleveland Clinic, and Stellantis. I also owned a Databricks platform with 600+ notebooks processing more than 454M profiles.

I moved the data-science insights pipeline to a Delta lakehouse, consolidating 22 jobs into a dependency-ordered Databricks Workflows DAG. I trained and served XGBoost and TensorFlow/Keras LSTM models, with batch inference over hundreds of millions of positions.

For Journey Builder, I built an LLM-driven pipeline that turns 50+ column spreadsheets into complete 73-task HR events in under four minutes. It routes across 40+ AI skills and uses GPT-4/5 and Claude Sonnet on AWS Bedrock.

At IBM Ireland, I worked on the Watson IoT platform and Spark ML pipelines for industrial analytics. My formal-methods work includes contributing to NASA Langley Research Center’s PVS libraries and formal verification and red-team analysis for the DARPA Voting System Demonstrator.