Jeremy Streett

Designed and shipped within 2 months of joining the org. Replaced the fragmented view landscape with a single daily ETL that reconciles 7 forecast types into consistent dimension mappings. Normalized historical forecasts that predated certain columns via CASE statement mapping so that every forecast (past and present) shares the same structure. Handles bundled vs. unbundled dimension variants with dedicated columns. The clean star-schema design is what makes everything else on this page work: the AI knowledge base references it directly so agents get correct answers on first query, the GTM Attainment App consumes it as its data source, and 5+ Tableau dashboards feed from it. Documented the full ETL logic in AtlasKB so any AI agent or analyst instantly understands how the dimensions map, what each forecast type represents, and how to query it correctly without guessing.

Self-contained analytics application: 2.2+ million fact rows loaded into the browser, every pivot/filter/aggregation executes client-side in under a second. Star-schema encoded integer arrays eliminate floating-point drift, reproducing source numbers to the unit with zero delta. Six interactive views including a pivot-anything crosstab (15 dimensions, 20+ metrics, 4 time grains), WBR calculated metrics, overlay charts, and forecast comparison. Formatted Excel exports with conditional formatting. Serverless pipeline (Spark → Lambda → S3 → CloudFront) auto-refreshes 3x daily.

Reverse-engineered the internal transfer portal API and solved the authentication challenge that blocked all prior tooling by reading Chrome session cookies via OS keychain decryption. Built a TypeScript MCP server with cross-platform support (macOS, Windows, Linux), auto-detection of Chrome profiles, and security hardening from day one: Zod validation, HTML sanitization, prompt injection prevention, response size caps, and structured logging with data redaction. First MCP server published to Amazon's official ASBX MCP Registry, with a pre-drafted Application Security Review covering threat model and mitigation evidence.

Designed, built, and operate a production BI-as-a-Service platform end-to-end. Officially approved by Intuit as a production QuickBooks application. Connects via OAuth, extracts financial data through CDC, computes KPIs (revenue, profitability, cash flow, seasonality), detects statistical anomalies, fetches external signals (weather, permits, housing data), and generates AI-powered weekly insight reports. Full ownership of every layer: Express/TypeScript backend, React admin UI, BullMQ pipeline, PostgreSQL via Prisma, Stripe billing, magic-link customer portal, and a cold outreach engine with self-built email infrastructure.