Built audience framework for a national furniture retailer incorporating lookalike modeling and clustering to define high, mid, and low value customer tiers across new and refined segments. Achieved 43% YoY ROAS improvement, 5M+ incremental reach, and 9% ROAS outperformance versus non-product campaigns.

Implemented custom bidding strategy for a national furniture retailer by identifying key optimization drivers — achieving 8% higher true reach and 30% more qualified prospects versus existing campaign benchmarks.

Engineered and productionized modeling and geolocation features into the blu. product, improving audience targeting and high-value geo identification. Drove 3–4× Revenue Per Impression improvement versus non-product campaigns across 25–30 client organizations, supported by technical documentation and training.

Developed ML-derived audience segmentation framework incorporating growth tiers and lookalike models, uncovering 10+ million net-new potential customers for national brands.

Built net-new audience spending and purchasing report benchmarking client customer behavior against the general population to inform campaign planning across 25+ clients representing hundreds of millions in media spend.

Self-initiated integration of third-party purchase and demographic data into audience products within two weeks, clearing 5+ pending client requests and accelerating campaign launch timelines.

Applied Claude Code across data engineering, feature development, QA, and UAT for AI/LLM-powered audience products — cutting 1–2 day turnarounds to a few hours, improving output completeness, and surfacing edge cases earlier.