I build the systems that decide what a distribution center or store should stock, and how much of it.
At O’Reilly Auto Parts I was one of two data scientists who created the assortment ML capability from scratch. Ours was among the first workloads in the company’s new GCP environment, and the tooling and pipeline patterns we established are what later teams built on. Those pipelines now refresh 200 models a quarter across 40 distribution centers, serve store-level assortments on weekly demand, and run against source tables holding tens of billions of rows.
I came to this deliberately — six years in financial services and education operations, then an MBA in Analytics and an MS in Data Science. That background is why I care as much about whether a merchandiser will actually trust and act on an output as about whether the model scores well.
Skills
4+ Years in the Data Science with an emphasis in supply chain.
4+ Years Experience: SQL, Python, BQ, BQML, VertexAI, DataForm, Snowflake, and Domo
2 Masters Degrees and Multiple Google Analytics Certifications.
Impact by the Numbers
A legacy forecasting process ran as a 21-hour manual Excel job. Migrating the logic to cloud-native Python and BigQuery brought it to five minutes, moving assortment decisions from a batch cycle to near real-time.
Scheduled and ad-hoc pipeline executions across 40 distribution centers. Five models per DC retrained, evaluated, and promoted with no manual intervention. A separate on-demand store program has served nearly 200 stores, fielding one to four requests a week.
In a pilot across four distribution centers, roughly 85% of the ~200 net-new SKUs the models surfaced went on to transact, giving merchandising a repeatable way to find products the manual process was missing.
Google Cloud Platform Experience
Data Form & Snowflake
Enterprise Dataform Orchestration: Implemented Dataform (SQLX) to manage complex dependency trees and data transformations, ensuring high-integrity model outputs.
Cross-Platform Integration: Streamlined the delivery of curated ML outputs from GCP into Snowflake, providing a centralized “Source of Truth” for organizational consumption.
Executive Visibility: Automated the data flow from 100s of sources into Domo, delivering real-time actionable insights to C-suite executives.
Big Query
In-Warehouse Machine Learning: Leveraged BQML to develop and deploy XGBoost propensity models and ARIMA demand forecasting directly within the data warehouse.
High-Volume Forecasting: Engineered scalable modeling frameworks to predict transaction likelihood and inventory depth for 200k+ SKUs across hundreds of locations.
Process Efficiency: Achieved a 99.52% runtime reduction by migrating legacy Excel-based calculations into highly optimized BigQuery SQL and BQML environments.
Agent Platform
Automated Performance Validation: Engineered individual Vertex AI pipelines for four distinct propensity models, incorporating automated evaluation against predefined success criteria.
Model Orchestration: Architected multi-stage GCP Pipelines to manage the end-to-end lifecycle of BQML models, from automated retraining to final output curation.
Hybrid Intelligence: Combined traditional ML (ARIMA/XGBoost) with cutting-edge Generative AI (Gemini 1.5 Pro) to solve complex business logic and narrative synthesis challenges.
Recent Projects
Warehouse Assortment Pipeline
Architected a sophisticated decision engine within the pipeline that automatically detects if a valid model exists for the current feature set.
Implemented a continuous evaluation loop for both new and existing models; pipelines utilize automated “Pass/Fail” logic to either promote a model to production or trigger an immediate error-log for manual intervention.
Developed a dual-path processing stream that combines propensity modeling with ABC/XYZ categorization, ensuring high-granularity demand forecasting across diverse SKU types.
Engineered automated selection logic that bifurcates SKUs into “Fast-Moving” (Automatic Pick) and “Slow-Moving” (Propensity-Driven) channels to optimize warehouse slotting and fulfillment efficiency.
Managed a multi-stage delivery system where finalized assortments are landed in Snowflake and visualized via Domo, facilitating warehouse-level implementation and real-time performance monitoring.
Tech Stack
- BigQuery & BQML
- Agent Platform
- Snowflake
- Domo
Ultrahub Assortment Pipeline
Developed a multi-layered ingestion engine that performs location and item clustering to transform store features, demand signals, and SKU universes into high-dimensional feature sets.
Engineered a dynamic decision engine that routes SKUs through specialized logic paths based on volume:
Fast-Moving: Automated selection paired with ARIMA forecasting for precision demand planning.
Slow-Moving: Selection driven by high-performance XGBoost propensity modeling.
Business Logic: Manual overrides for strategic SKU placement and specialized constraints.
Built automated logic to calculate inventory depth adjustments post-selection, ensuring that the final assortment is optimized for both variety and availability.
Architected an end-to-end feedback loop where model outputs land in Snowflake for Domo dashboarding, allowing for stakeholder evaluation and store implementation before re-initiating the pipeline based on new stakeholder requests.
Tech Stack
- BigQuery & BQML
- Agent Platform
- Snowflake
- Domo
D&D Recap | Multimodal Narrative Intelligence Pipeline
Engineered a custom pipeline to process multi-channel Discord audio using OpenAI Whisper, implementing speaker identification and diarization to convert raw audio into structured scripts.
Leveraged Gemini 1.5 Pro’s long-context window to ingest extensive session scripts, utilizing advanced prompt engineering to generate consistent narrative recaps and stylized “narrator-perspective” scripts.
Integrated ElevenLabs API with a custom-trained voice clone to automate the production of high-fidelity audio recaps, delivering a professional-grade narrative experience.
Built the end-to-end workflow to handle the transition from unstructured audio data to polished, multi-format (text and audio) creative assets.
Tech Stack
- Discord
- Whisper
- Gemini
- ElevenLabs
Contact
Please reach out to me via Linkedin with any questions or job opportunities!
