Carlitos Willis

Los Angeles, CA

• Built data aggregation tools in Python and Node.js to track audience growth and engagement across Spotify, TikTok, and YouTube, consolidating data into PostgreSQL.
• Developed Selenium-based web scrapers and API integrations to ingest 5K–20K+ data points daily, normalizing inconsistent schemas across platforms.
• Designed automated data pipelines to process streaming and social data, generating growth signals used to evaluate dozens of emerging recording artists.
• Delivered data-driven insights that reduced manual analysis time and informed audience targeting, engagement strategy, and release planning for music industry stakeholders.

Los Angeles, CA

• Designed and executed dozens of A/B tests with React, TypeScript, and Ruby — nine in just six months, achieving a 15.5% increase in Visit-to-Lead Conversion Rate (CvR) and a 5.8% year-over-year increase in Direct 30-day Lead-to-Join CvR.
• Led the redesign of the experiment tracker and setup process, significantly cutting setup time by 40–60% and increasing process transparency.
• Developed an in-house testing hub, reducing experiment-sharing setup from an hour to 5 minutes.

NYC + LA

• Automated content planning, creation, and marketing workflows using Python scripts to fill templates, enhancing operational efficiency and reducing onboarding time.
• Overhauled lead gen for businesses and recording artists through landing pages with tracking pixels.

Berkeley, CA

• Expanded social media reach 10x with targeted advertising, analytics, and custom Python scripts.

Working Memory — Event-sourced task board you can rewind to any past state

Next.js, TypeScript, React, SQLite, Litestream, Backblaze B2, Docker, dnd-kit, Tailwind

• Built an append-only, event-sourced task board where every edit, move, and completion appends to an item_events log via SQLite triggers — making the board’s full state reconstructable at any past point in time, with multi-account isolation enforced at the query layer.
• Deployed on Render with a self-healing durability pipeline: Litestream streams the database to Backblaze B2 and restores it on every boot, backed by daily automated pulls and an uptime worker that eliminates cold-start churn.

Ableton Library — Rust catalog + desktop app for a decade of Ableton projects

Rust, SQLite FTS5, Tauri, TypeScript

• Wrote a lenient Rust parser for Ableton’s undocumented .als format (gzipped XML) feeding a SQLite FTS5 catalog: search hundreds of projects by tempo, plugin, sample, or artist in milliseconds, with waveform previews and a 3D similarity map in a Tauri app.

OP-1 Lab — Firmware reverse engineering; custom DSP on real hardware

Blackfin assembly, Python, GNU bfin toolchain, op1repacker

• Mapped the OP-1 synthesizer’s closed Blackfin firmware and injected a hand-written assembly synth engine that runs on the physical instrument — byte-verified patches, a host-side Blackfin interpreter for pre-flash verification, and strictly reversible changes.

armak.tech — Production website for a low-voltage & networking firm

Vite, EJS, Vanilla JS, HTML Canvas, Web3Forms

• Built and shipped the live production site for a real business: a static Vite + EJS build with hand-written HTML Canvas animations (a mouse-reactive particle network, an animated SVG service-area map) lazy-loaded per viewport via IntersectionObserver, a backend-less Web3Forms contact form, and full Open Graph / Twitter SEO metadata — cutting Largest Contentful Paint ~90% by serving WebP heroes with preload and fetchpriority.

Ezzy Shop — E-commerce service split into independently-scaled microservices

AWS EC2, Nginx, Redis, PostgreSQL, Node.js, Express, React, Docker, K6, Loader.io

• Boosted load capacity from 200 to 800 requests/sec and cut server response time from 230ms to 70ms by load-balancing 5 AWS EC2 instances behind Nginx with Redis caching and remedying database-schema bottlenecks.