Prospexa

Business Intelligence Analyst, AI Agent Development

Jan 2025 - Present • 1 yr 7 mos

Builds AI agents in Snowflake that let business teams ask questions of their own data directly, instead of every request turning into a ticket that waits three days for an analyst. The unglamorous part is what makes those agents right. Defining what each business term actually means, classifying data assets, documenting the rules in Ataccama, and watching data quality across Consumer BI every day. An AI agent can only answer a question correctly if someone already decided what the metric means and where the real number lives. Almost every AI project skips that step, which is exactly why those tools answer confidently and answer wrong. The rest of the role is the reporting that a large credit union runs on. Tableau dashboards built on Snowflake and Tableau Prep covering customer, product, servicing, and operational metrics, in front of stakeholders who will notice immediately if a number moves without a reason. Works alongside BI developers on ETL and reporting pipelines, tracks down where data breaks, and delivers in an Agile cycle using JIRA and Snowflake Notebooks. Working in a regulated financial environment sets the bar for how the rest of our builds get done. Numbers get validated before anyone sees them, and every metric has an owner.

Research Data Analyst, Automation and Reporting

Feb 2024 - Present • 2 yrs 6 mos

Benchmarking research programs running across hundreds of participating companies at once, where the same collection, scoring, and reporting cycle repeats every program. Most of that cycle was being done by hand when I got there. I automated it. Survey and collection. Built the survey infrastructure and SQL ingestion pipelines that pulled responses in as they arrived, with validation logic catching incomplete and inconsistent submissions before they ever reached analysis instead of after someone spotted a bad number in a client report. Data pulls and processing. Wrote R pipelines that cleaned and standardized high volume survey data across wildly different company sizes, industries, and response formats. Once it was built, each program cycle ran on the same code instead of getting rebuilt from scratch by whoever was assigned to it. Scoring and reporting. Replaced manual scoring, weighting, and aggregation with automated processing, work that used to take days per program. Built ranking models and performance tiers in SQL and R, feeding publication ready benchmark reports and Power BI dashboards that went straight into client presentations. HubSpot and ticketing. Automated the ticketing and workflow side so participant questions, submission issues, and program communications got routed and tracked automatically rather than living in someone's inbox. These were commercial products, not internal reports. The dashboards and publications were how the company sold and retained clients across multiple industry verticals, so the automation was not a nice to have. It was the delivery capacity.