Caleb Dame

Serve as primary technical point of contact for a Tier 1 U.S. bank, supporting AI-driven sanctions and PEP/adverse media screening models within its financial crimes compliance program.

Lead improvements to model interpretability, improving reason codes derived from SHAP attributions into clear, actionable, and logical explanations for investigative partners and internal compliance.

Author model documentation and partner closely with the client's internal Model Risk Management team through internal validation, presenting and defending methodology to meet regulatory and compliance standards.

Designed predictive models for strategic product marketing, churn prediction, and collections optimization used by credit unions and utility partners, directly impacting program adoption rates and retention strategies.

Developed and maintained privacy-affirming federated prediction models, dynamically aggregating insights from internal data to develop dynamic, industry-wide product models while protecting client data privacy.

Built scalable internal tooling to standardize data science processes, streamline project throughput, monitor model performance, and expose model interpretability clearly to end users.

Built ML-driven lead qualification models that improved sales efficiency by eliminating 20–40% of low-quality meetings and increased top salesperson conversion rates 2–3x.

Designed and managed ongoing A/B testing programs to optimize marketing funnel performance and application outcomes.

Created and oversaw CRM integrations (HubSpot, Close, Zapier) and real-time dashboards to enable data-driven marketing decisions.

Shipped fraud detection models using LightGBM and XGBoost, and maintained internal packages for measuring drift in fraud typologies with long maturity windows and estimating changes in risk prevalence over time in unlabeled data.

Implemented and benchmarked advanced ML architectures (Autoencoders, Stacked Models, and Graph Neural Networks) to research applications for identifying fraud networks.

Led cross-team collaborations over the course of an acquisition, integrating newly acquired data assets to enrich features and model performance.

Designed and maintained 15+ ETL pipelines in Databricks, Airflow, and Snowflake, processing 3.5 TB of transformed data daily for financial analysts.

Built automated anomaly detection using clustering algorithms, increasing data purity by 10–15% and enabling real-time visualization in Tableau dashboards.

Researched ML applications (XGBoost, Neural Networks, Naive Bayes) to extract insights from large-scale census data.

Managed large-scale databases in MS SQL Server, ultimately reducing record-linking error by 80%.