Hollis Cuffie

Hollis Cuffie

Senior Data Scientist building production ML systems and causal frameworks that drive tens of millions in impact.

LinkedIn GitHub Download CV

About Me

Senior Data Scientist with 5+ years building production ML systems and analytical frameworks at the intersection of product, experimentation, and AI. I work end-to-end — from hypothesis to model to stakeholder decision — and care as much about whether the insight gets used as whether the math is right.

My Journey

P&G

Senior Data Scientist

Jul 2024 – Present

Owns the ML platform behind $300M in revenue decisions. Re-architected core models, deployed production RAG systems, and built causal frameworks that captured $20M in whitespace opportunities.

Enterprise ML Platform Causal Inference Time-Series Forecasting RAG Systems
P&G

Embedded Lead Data Scientist

Aug 2022 – Jul 2024

Led market-scale causal experimentation across 20+ deployments. Built P&G's first production GenAI application and the observability infrastructure to keep it honest.

Market-Scale Experimentation GenAI LLM LLM Observability
P&G

Analytics & Insights Manager

Aug 2020 – Aug 2022

Owned product analytics for P&G Beauty's iOS app end-to-end — instrumentation, A/B testing, behavioral cohorts, and weekly executive readouts shaping roadmap and spend decisions.

Product Analytics A/B Testing Behavioral Modeling
🐘

Business Analyst, Product Team

One Kings Lane • May 2019 – Apr 2020

Supported migration to an ML-driven search relevance engine, applied topic modeling to unstructured returns data, and built vendor scorecards identifying $250K in preventable losses.

Mobile App Analytics ML Search Optimization Topic Modeling

Skills & Tools

Python SQL PySpark Causal Inference A/B Testing CUPED Uplift Modeling Switchback Designs DiD Synthetic Control Survival Analysis XGBoost SHAP Regression Clustering & Segmentation Time-Series Forecasting Python SQL PySpark Causal Inference A/B Testing CUPED Uplift Modeling Switchback Designs DiD Synthetic Control Survival Analysis XGBoost SHAP Regression Clustering & Segmentation Time-Series Forecasting
RAG Systems LLM Evaluation LangChain / LangGraph Embeddings Hallucination Detection Prompt Engineering OpenAI APIs Databricks GCP / BigQuery Azure dbt Product Analytics KPI Design Funnel & Cohort Analysis LTV / CAC / Churn Executive Communication RAG Systems LLM Evaluation LangChain / LangGraph Embeddings Hallucination Detection Prompt Engineering OpenAI APIs Databricks GCP / BigQuery Azure dbt Product Analytics KPI Design Funnel & Cohort Analysis LTV / CAC / Churn Executive Communication
Observational Causal Inference Geospatial Modeling Mobile Behavioral Modeling Propensity Scoring Simulation Ragas Arize Cross-Device Modeling Metric Pipelines Batch & Near-Realtime Scoring Model Monitoring Regression Testing Roadmap Influence Cross-Functional Leadership Platform Measurement Strategy Observational Causal Inference Geospatial Modeling Mobile Behavioral Modeling Propensity Scoring Simulation Ragas Arize Cross-Device Modeling Metric Pipelines Batch & Near-Realtime Scoring Model Monitoring Regression Testing Roadmap Influence Cross-Functional Leadership Platform Measurement Strategy

Projects

🎯

Uplift Modeling for Churn Intervention

Predicting who will churn isn't the same as predicting who will respond to intervention. Uses meta-learner uplift models on the Criteo dataset to identify persuadable customers — and quantify the ROI of targeting them vs. the full population.

Uplift Modeling Causal Inference Qini Curves Python
📊

MarketplaceCI In Progress

Causal evaluation of dynamic pricing policy impact in a two-sided marketplace. Uses DiD, synthetic control, and spillover modeling on public Airbnb NYC data to simulate real-world policy tradeoffs.

Diff-in-Diff Synthetic Control Spillover Modeling Python
🔍

Custom SHAP Interpretability Toolkit

Reusable SHAP waterfall and driver-analysis tooling built for production propensity models. Makes model explanations legible to non-technical stakeholders without sacrificing rigor.

SHAP XGBoost ML Interpretability Python
🚧

More coming soon

RAG evaluation, LTV modeling, and more in progress

Beyond the Code

Hollis

Abolition

Animal Rescue

Arabic + Japanese

Art

Taiko

Good Food

Metaverse

NBC Sing-Off

Rare Plants

Responsible AI

Travel

Volunteering

Word Games

Snorkeling

Hiking

Anime

Volleyball

Fencing

Duolingo

Reproductive Justice

Karaoke

Skiing

Electric Guitar

Origami

Pickleball

DIY Projects

Mixology

Pilates

Mental Health

Board Games

Get in Touch

I'm always open to discussing new projects, creative ideas, or opportunities in data science and beyond. Feel free to reach out!