Ibuilddatasystemsthatscale. Full stop.

Health & Activity Analytics Pipeline

Designed a production-grade data platform architecture around asset-centric orchestration, then directed AI tooling to build within it. I architected the medallion model (Bronze/Silver/Gold), defined data quality contracts via Great Expectations, and established Kafka streaming patterns. The result: a documented human-in-the-loop workflow where architectural decisions are mine and mechanical implementation is AI-directed. 17 passing dbt tests, 3 validated GX suites, real-time streaming extension via Kafka.

graph LR; Oura/GitHub/OpenWeatherMap -> Dagster Orchestration -> DuckDB Bronze -> Silver -> Gold -> dbt Core Transforms -> Streamlit Dashboard.

Email Service Provider Migration

The Philadelphia Inquirer needed to exit Salesforce Marketing Cloud without losing subscriber history or segmentation, so I designed a controlled migration path. I built pipelines from Salesforce Service Cloud into Marigold SailThru with rigorous field mapping, validation, and reconciliation. The cutover preserved segmentation and engagement history, achieved zero data loss, and reduced annual platform costs by over $100K.

graph LR; Salesforce Marketing Cloud -(Legacy State)-> Migration Pipelines; Salesforce Service Cloud -(Subscriber Records)-> Migration Pipelines -> Field Mapping + Validation -> Reconciliation Layer -> Marigold SailThru -> Segmented Lists and Engagement History.

Enterprise production system

Serverless ETL Platform

Legacy billing ETL ran on always-on servers and required frequent manual intervention, so I replaced it with an event-driven AWS architecture. The flow used EventBridge, Lambda, SQS, DynamoDB, and S3 to decouple ingestion, transformation, and reconciliation while maintaining state and auditability. This shift cut infrastructure costs by 75% and reduced manual operations by half.

graph LR; Billing Event Trigger -> EventBridge -> Lambda Ingest -> SQS Queue -> Lambda Transform -> S3 Raw Storage and DynamoDB State + Audit -> Lambda Reconcile.

Enterprise production system

AI/RAG Historical Archive

The Inquirer's 45-year archive was locked in siloed content systems with no unified search capability, so I designed and deployed a privately-hosted RAG application to surface it. I built an embedding pipeline over 1977-to-present articles, set up a vector retrieval layer, and hosted the inference stack internally to satisfy legal and editorial controls around proprietary content. The result: journalists and editors can now surface decades of institutional knowledge in seconds.

Enterprise production system

ETL Pipeline Reengineering

Legacy Informatica pipelines were brittle, expensive to maintain, and locked behind proprietary licensing that inflated costs without delivering flexibility. I rewrote the critical paths in Python and PySpark, replacing vendor-owned processes with testable, version-controlled code. The migration delivered 50% better resilience on failure recovery and eliminated the Informatica licensing overhead entirely.

Enterprise production system

Internal HR Chatbot

HR was fielding repetitive questions about PTO, org structure, and policy that consumed significant time and created inconsistent answers. I built a RAG-powered chatbot embedded directly in Slack, indexing the employee handbook, org charts, and policy documents into a vector store backed by the OpenAI API. Employees get authoritative, instant answers without leaving their workflow; HR gets fewer tickets.

Enterprise production system

Full Stack GRC Platform

Longevity Consulting's GRC processes were entirely manual, tracked across spreadsheets with no audit trail, no workflow automation, and no visibility across risk domains. I designed and built a full-stack platform with a React front end, Django and FastAPI services, a GraphQL API layer, and containerized deployment on AWS Fargate. The platform gave compliance teams a single source of truth with automated workflows and full audit history.

Enterprise production system