Projects.

Built a Python RAG assistant answering natural-language questions over 1,000+ transit documents using a vector store for fast semantic retrieval. Orchestrated a 4-step agent with LangChain and LangGraph, retrieving context then calling the Claude API for grounded, cited responses. Exposed 3 retrieval and live-data tools through MCP for reusable tool integration, and evaluated retrieval across 100+ test queries to tune chunking, prompts, and embeddings.

Engineered an ETL pipeline ingesting 100k+ public transit records from REST APIs into Databricks on a scheduled daily batch load. Modeled data across 3 medallion (bronze, silver, gold) layers, using SQL and PySpark to transform raw API responses into clean, usable tables. Implemented 15+ data-quality checks and schema enforcement across layers to preserve a reliable single source of truth for downstream analysis, and automated testing and deployment through a CI/CD pipeline across 2 environments.

Built a Dual-Shore sourcing model for a $48M Gen Z apparel brand under tariff pressure, shifting 30% of production to Mexico to reclaim ~$2.40/unit versus Asia sourcing. Modeled cost-benefit tradeoffs across Asia, Mexico, and U.S. lanes, quantifying tariffs and freight to defend margin, then paired a 10–12% sustainability-anchored price lift with an 18-month roadmap scaling suppliers from 3 to 8+.

Engineered a 15-table Oracle relational database modeling a complex business operation, decomposing 5 many-to-many relationships into junction tables to enforce data integrity. Authored 40+ SQL queries using multi-table joins, aggregates, subqueries, and set operations, and built a recursive CONNECT BY query to generate an indented org chart. Solved several problems three different ways on purpose, just to learn which approach fits where.

Led an 8-person cross-functional team, the largest I'd led at the time, in a Shell-judged competition comparing Vaseline against shea and cacao butter alternatives on cost, sustainability, and safety. Split the team into three sub-workstreams so decisions didn't bottleneck through me, and crafted the go-to-market transition framework tying the three pillars together.

Given a mid-market client facing a ~35% retention decline and a 48-hour deadline, diagnosed root causes via Python, SQL, and Power BI, then designed a $150K technology modernization roadmap prioritizing CRM, automated analytics, and process digitization. As Milestones Lead, broke the 48 hours into checkpointed sprints so the deck never slipped.

Directed a 4-person interdisciplinary team through a structured case engagement, setting workstreams to deliver client-ready work by deadline. Cleansed and consolidated client datasets in Excel, eliminating 100% of analytical noise to reframe the problem and sharpen strategic insight, then formulated a supply chain optimization strategy via root-cause analysis, projecting 35% in cost savings through a phased deployment roadmap.