Staff Engineer Machine Learning
CurrentDesigned and built the primary backend systems that are used to manage incoming streamed event data that relays emerging facts about campaigns and supporting state from Amazon Ads and SP Apis. The system uses an event sourced data model that allows for domain driven design modeling to be achieved.Designed and built the primary system used to interact with OpenAI and other LLMs to allow chat based interactions with Snowflake fact tables containing performance and state data. The chat system leans on use of an Agentic graph that allows the LLM to decide next steps and retries those options if the results are less than ideal. This opens the door to some very elaborate ad hoc queries that are able to answer performance questions that might have taken months to plan and build custom UIs for. These interactions allow for answers to be obtained in moments using a normal conversational style.