About Me
Data leader with a passion for turning insights into impact
I build things that matter at the intersection of data and music. Right now, I'm the founder of Sync Wave Analytics, where I'm creating the analytics infrastructure that helps labels, investors, and artists understand what their catalogs are actually worth—processing 500K+ daily streaming events across Spotify, Deezer, and Luminate.
Before going all-in on music tech, I spent five years at Spotify building the systems behind the scenes—churn prediction models serving 500M users, the data science for Wrapped campaigns, and ML features like Family Mix that shipped to millions. Earlier, I detected fraud patterns in 10M+ daily transactions at JPMorgan Chase (and earned a utility patent for privacy-preserving analytics), and cut my teeth building fantasy sports models at ESPN that powered content for 2M+ weekly users.
I studied Mathematics and Statistics at Rutgers, but what really drives me is turning messy, complex data into something that helps people make better decisions. Whether that's a $100M acquisition call or figuring out which songs deserve a second listen—I love the puzzle.
Experience
From sports analytics to streaming at scale to building my own thing
S
- Founded Lyric Nexus, a B2B music analytics platform processing 500K+ daily streaming events from Spotify, Deezer, and Luminate to fuel catalog valuations, royalty forecasting, and artist performance insights for labels and investors.
- Architected Snowflake data warehouse with 260+ dbt models and automated Slidev decks, cutting analyst report generation time by 80% while keeping output reproducible and tailored to every stakeholder, from investors to artists.
- Trained a content-based recommendation engine blending audio feature embeddings and genre similarity scoring, hitting 89% precision on sync licensing matches and supporting playlist curation for catalog discovery.
- Engineered full-stack web app with Next.js, TypeScript, and Neon PostgreSQL; implemented OAuth 2.0, RBAC, and row-level security for multi-tenant access across 15+ enterprise clients on a SOC 2-ready architecture.
- Orchestrated ETL pipelines with Prefect and Airbyte pulling royalty feeds from 20+ streaming services, handling 250GB+ each month at 99.5% reliability with sub-hour freshness for real-time analytics.
- Built reusable post-analysis templates measuring A/B testing lift from ranking and retrieval updates to Disney streaming recommendation systems, speeding experimentation readouts, raising analytical rigor, and freeing up capacity.
- Deployed the team's first Snowflake Cortex agent and Databricks Streamlit apps wired to Jira/Confluence MCP servers via Claude Code and Cursor, giving Disney Personalization one place for resources and observability.
- Pioneered AI adoption at Disney as a founding member of the org's AI Charter, working cross-functionally to democratize AI for hundreds of employees; coached teammates and presented at company forums.
- Introduced dbt best practices into Disney's analysis workflows, aligning with upstream data engineering so stakeholders got accurate numbers; onboarded experimentation, query-builder, and BI tools at record pace.
- Directed a 5-person data science and analytics group for a PE firm buying music catalogs, producing valuation models and due diligence datasets that backed $100M+ in acquisitions and secured 3 investor funding rounds.
- Migrated legacy infrastructure to GCP with serverless ETL pipelines moving 250GB+ of transaction data from 200+ royalty sources, lowering processing costs by 60% and widening cross-team data access.
- Launched a Centralized Insights Hub joining Spotify, Deezer, and Tidal streaming data with catalog financials, unlocking track-level revenue attribution and shrinking monthly reporting cycles from 2 weeks to 2 days.
- Designed an end-to-end churn prediction pipeline scaled to 500M users on TensorFlow and BigQuery, yielding 10M reactivations and $20M in incremental revenue through targeted re-engagement campaigns.
- Led cross-functional data science for the Spotify + Hulu Bundle launch; propensity models and A/B testing frameworks sharpened targeting and messaging, adding $15M in paid subscriptions in its first quarter.
- Managed the DS and Analytics Engineering team for Wrapped 2019-2020, Spotify's most viral annual campaign; brought in a matched-market methodology that quantified 1M incremental listening hours and proved its ROI.
- Drove Product Data Science for a new premium tier and the Spotify Stations app, running experiments, UX research, and growth strategy that shaped roadmap decisions for 50M+ monthly active users.
- Shipped two ML personalization features to production for millions of listeners: "Family Mix" (the first cross-account algorithmic playlist) and emotion-based recommendations from image recognition.
- Delivered NLP and ML systems classifying 10M+ credit card transactions per day for merchant matching and bot detection at 92% accuracy, trimming fraud investigation time by 40%.
- Granted utility patent on privacy-preserving synthetic data methods for secure cross-institutional analytics.
- Created a predictive sports analytics suite for Daily Fantasy, with ML lineup optimization across NBA, MLB, and NHL reaching a 73% top-quartile hit rate and feeding content for 2M+ weekly readers.
- Developed Monte Carlo simulations for MLB run expectancies, 85% accurate and featured on ESPN broadcasts.
Projects
Personal projects showcasing full-stack development and data engineering
Applications
Full-stack web applications built for scale
An invite-only family platform that puts an Obsidian-style wiki, blind ranking and tier-list games, group chat with an AI member, a Spotify player, Google Calendar, stock and crypto dashboards with a NY Fed recession model, and flood-risk reports built on FEMA and USGS data behind one hardened login. AI agents edit the wiki and play the games through three OAuth-protected MCP servers, and the same assistant answers in Discord. It also runs the family homelab: Home Assistant managed as code with a heartbeat, leak alerts, and locked offsite backups, plus a Proxmox host running Pi-hole.
A retro 8-bit music guessing game with three modes: a Daily famous song whose decade is revealed only after you solve it, Classic rounds from any artist, genre radio, or chart, and Soundtracks, where you hear licensed songs and name the video game. Clips grow from 2 to 30 seconds over 5 attempts, and players can start as guests, build streaks, climb global and friends leaderboards, and share result grids. A TanStack Start app and an OpenAPI Deezer proxy run on Cloudflare Workers, backed by Neon Postgres.
A real-time fantasy sports site for leagues with friends, with rules calibrated on 70+ real Yahoo league setups: drafts with a value-over-replacement board, mock drafts against bots, lineups, trades with league votes, three waiver systems, and a commissioner suite with an audit feed. Bots can take empty seats, lineups optimize exactly, standings track clinch and elimination magic numbers, and live scoring pushes to every open page. Launching with NHL leagues for the 2026-27 season.
Baseball research built on a dbt and BigQuery warehouse: a Baseball Savant-style explorer with filterable hexbin strike zones, spray charts, and four-player comparisons in shareable URLs, a first-principles hitter WAR model, and insight pages on playoff predictability, busts and sleepers, hitting streaks, pitcher durability, and 25 years of drama-scored games. Findings are presented as Slidev decks with Vue 3, D3, and ECharts visualizations.
Data Engineering
Data pipelines, warehouses, and APIs
Tracks jersigamis: jersey-number combinations from a single play that have never happened before, like a batter plus the fielders who handled the ball, or an NHL goal scorer plus assists. The site pairs a live feed with leaderboards, team standings, team and player dashboards, NHL goal replays, and an "Are we running out?" look at how novelty has changed since 1917. A Cloud Run watcher checks live games every minute and scores each new combination with an information-theoretic rarity model, while a nightly Bruin pipeline rescores full history across DuckDB, BigQuery, and Postgres.
Technical Skills
Technologies and tools I work with
AI & ML
Backend & Data
Infrastructure
Languages
Analytics
Frontend
