Selected work · A journal

The stuff I'm proud of.

I'm Cort — a software engineer who ships. Real-time audio engines in C++, AI-native applications, and data systems that turn mess into something you can actually use.

25+ years building for Fortune 500s and startups across finance, energy, healthcare, and marketing. This page is the highlight reel — the projects that show what I do when I get to build the whole thing.

0
Years shipping software
0
Industries delivered in
0
Synth presets crafted
0
Pages auto-generated

The highlight reel

Things I built — and why they matter

Each of these I designed and built end to end: the DSP or the data pipeline, the AI integration, the interface, and the page selling it. Newest first.

September 2026 · Live on its own domain

Corticorp Finance

Live
The Corticorp Finance strategies table with a period selector and a vs S&P 500 verdict on every row

The most ambitious thing I have built. Thirty-nine simulated portfolios run side by side on the same live market, each with its own stated investment discipline, each starting from the same $100,000, and each publishing every decision it makes along with the reasoning and the exact inputs it saw. It began as a tab inside Corticorp News and in late August 2026 it moved out to its own domain, its own masthead and its own five sections. Simulated, and not investment advice; the point is the experiment, not a tip.

  • Thirteen model-driven strategists plus twenty-six documented sleeve portfolios, including two deliberate control arms that exist so the roster has something honest to be measured against
  • An AI Board reviewing the experiment in public: a Judge, a Scientist, a Treasurer, an Auditor, an Editor, a research Librarian and a Marketer, with their findings and open work items published rather than kept internal
  • Delta-1 risk telemetry: every decision carries a snapshot of what a 1, 3, 5 or 10 percent market fall does to that exact book, computed by revaluing each position rather than by a rule of thumb
  • A written register of every change to the models behind it, because changing a model changes the measuring instrument and a track record that spans one is two records, not one
  • Two products from one codebase, routed by hostname, each domain publishing only its own pages, sitemap and robots file
Claude APINext.js 16SQLiteRailway
Project page & diary → Open live →
2026 · Live, running on its own

Corticorp News

Live
Corticorp News Situation Room satellite map with military symbology and a front-line overlay

An AI-curated digest that reads dozens of paywall-free outlets and gives you one synthesized summary per real-world event instead of ten near-duplicate headlines. What I am proudest of here is the part that admits doubt: a public Verification Desk that checks circulating claims against independent reporting and is willing to publish "unverified" rather than smooth it over.

  • Articles clustered into real-world events, then summarized once, neutrally, with the sources shown
  • Map-driven situation rooms for ongoing conflicts, with corroborated public event locations and a front-line overlay
  • Scheduled digest runs inside the server process, so the feed refreshes itself with no separate worker to babysit
  • Full-text search across every event ever published, with extracted article imagery archived alongside
Claude APINext.js 16SQLiteLeaflet / Maps
Project page → Open live →
2026 · Live

World’s Eye View

Live
World's Eye View at continental zoom, live webcam thumbnails scattered across a satellite map

Around 34,700 public webcams from 31 open sources, quilted onto one satellite map as live thumbnails rather than pins. Zoom out and you get the notable ones; zoom in and a city fills with windows. Because every frame is real and refreshes on its own schedule, the map drifts through the day with the world, and you can watch dawn cross a continent.

  • Server-side thinning that projects every candidate into screen pixels and keeps the most interesting camera per grid cell, which is what stops 4,800 Florida cameras becoming a solid wall
  • Frames proxied and downscaled to roughly 6KB each, so one origin serves every image and the security policy never has to trust dozens of third-party hosts
  • A multicam wall, cameras along a driving route, and a finder for wherever sunrise or sunset is happening right now
  • The map needs no database at all; the one behind it exists only for optional accounts and saved walls, and the map keeps working without it
Next.js 16LeafletEsri imagerysharp
Project page → Open live →
2026 · Live

AlgoRythmic

Live
A plain-English description compiles to editable rules, which generate a queue you preview for free, then save to YouTube Music

A playlist builder for YouTube Music that treats a playlist as a rule set you can edit. Describe a vibe in plain English and it compiles to visible selection rules: era, length, no live versions, never the same artist twice. If the queue is wrong, you can see which rule made it wrong, which is the thing no streaming service will let you do.

  • Built around one hard constraint: the API gives 10,000 quota units a day and a 50-track save costs about 2,550, so everything except saving runs free against a local mirror of your library
  • Seed a queue from one song and weight eight dimensions of what "similar" is allowed to mean
  • Every unit spent is written to a ledger, and a call that would blow the day's budget is refused rather than failing halfway through a save
  • Saves a real private playlist, so it plays in the car with background playback and lock-screen controls, which an embedded player cannot do
Claude APINext.jsPostgresYouTube Data API
Project page → Open live →
2026 · In the lab

CortComp

Prototype
Input level → Output level → 1:1 Threshold
Upward (softs ↑) Downward (louds ↓) Invert (beyond ∞:1)

A dynamics processor that does something most compressors physically can't: make the softs loud and the louds soft — past the point of a limiter. Its INVERT mode drives the transfer slope negative (beyond ∞:1), so louder input yields quieter output: true dynamic inversion.

  • Header-only, JUCE-free DSP engine — unit-testable offline, no framework lock-in
  • Stereo-linked feedforward detector that reacts to the louder channel to preserve the stereo image
  • Three simultaneous behaviors — downward, upward, and inversion — from one clean transfer curve
C++ / JUCEDSPVST3Real-time
2025–2026 · Shipped & free

CortSynth

Shipped
CortSynth main synthesizer interface

A complete VST3 synth, sampler, and drum machine in one plugin — the "everything" instrument. Dual-oscillator hybrid engine, physically-modeled instruments, 200+ presets, and an AI that builds patches from a plain-English sentence. Downloadable, with a demo video and real users.

  • AI Tone Designer: describe a sound in words, get an editable patch — Claude API driving real synthesis
  • Physical modeling: Karplus-Strong strings, a dispersive stiff-string piano with resonating soundboard and sympathetic strings
  • Zero heap allocations on the audio thread — glitch-free under load, built for live performance
  • 2× oversampling, FDN reverb with shimmer, a rotary Leslie, and an offline preset generator
Modulation Constellation
Sampler mode
Drum pad grid
C++ / JUCEClaude APIDSPVST3 + Standalone
Product page → Demo video →
2026 · Commercial release

CortVerb

$99 · Launching
CortVerb reverb interface with live visualization screen

A commercial VST3 reverb built on a Hybrid FDN engine, wrapped in a hardware-grade interface: a central digital screen that visualizes the reverb live across six pages, an auto-duck sidechain, and 12 tactile knobs. Proof I can take an engine all the way to a product people pay for.

  • Hybrid feedback-delay-network reverb tuned for musical, mix-ready tails
  • Real-time visualization screen with six views — the reverb, made legible
  • Auto-duck sidechain so the reverb ceilings gracefully under a busy mix
C++ / JUCEDSPVST3Product
Product page →
2026 · Live demo

Apartment Finder

Live
Apartment Finder dashboard with interactive map and scored listings

An AI-native rental aggregator that scrapes listings from six sources, then uses vision AI to actually look at every photo — scoring quality and reading off flooring, natural light, and kitchen style. Everything lands on an interactive map-and-table dashboard with commute scoring.

  • Multi-source scraping across Craigslist, Zillow, Apartments.com, Trulia, Realtor.com, and Zumper (Playwright)
  • Gemini Vision scores photo quality and extracts features no listing field contains
  • Auto-generated 450+ individual detail pages with galleries, lightboxes, and commute breakdowns
  • Scam detection from price anomalies and stock-photo patterns; work-commute distance scoring
Gemini VisionPythonPlaywrightLeaflet / Maps
Project page → Live demo →
2026 · Live tool

CatfishCheck

Live
CatfishCheck investigation interface

A self-service OSINT tool for verifying online identities. Upload photos, phone numbers, social handles, and messages about a suspected scammer; get back a report that surfaces the red flags — built with privacy as a first principle.

  • Reverse image search, carrier lookup, and a geo-verification beacon in one workflow
  • AI-assisted analysis that turns scattered evidence into a readable risk report
  • Case-based investigation UI that keeps each subject organized and reviewable
AI AnalysisJavaScriptOSINTPrivacy
Project page →
2026 · In development

Ironclad

Coming soon
Ironclad amp-style distortion plugin interface

An amp-style distortion plugin with four distinct voicings and a focused tone stack, dressed in a bold hardware-inspired UI. Rounds out a coherent plugin line — synth, reverb, distortion, compressor — all sharing an engineering discipline.

C++ / JUCEDSPVST3Distortion
Product page →

Why hire me

What I actually bring to a team

The through-line across everything above: I take an idea from a napkin to a shipped, polished thing — and I use AI to move faster than a team without it.

🚀

I ship the whole thing

Engine, interface, and the page that sells it. I don't need a spec handed down in three parts — I can own a product from architecture to launch.

⚡

Hard real-time engineering

Zero-allocation audio threads, oversampling, physical modeling, DSP that can't glitch. If a domain has tight constraints and no room for error, I'm comfortable there.

🧠

AI where it does real work

Not a chatbot bolted on. Vision models that read photos, language models that generate valid patches, pipelines that score and decide — AI as a load-bearing part of the product.

🗺️

Mess into something usable

Scraping, cleaning, enriching, and visualizing real-world data — turning unstructured listings, photos, and text into searchable, scored, actionable dashboards.

The toolkit

C++PythonJavaScript JUCE / DSPClaude APIGemini Vision FastAPIReactNode.js Playwright / ScrapingSQLAWS DockerLeaflet / MapsGit

About this page

This page is itself the argument. I built it with an AI coding agent as a pair — it read my whole body of work, matched my site's design system, and helped me shape and polish this in a single sitting.

That's the point I want a future employer to take away: I don't just use AI as a feature in products — I work with it as a force multiplier. The same leverage that put this page together is the leverage I bring to a team's roadmap. More shipped, faster, without cutting the quality corner.

Let's talk

Want someone who ships?

If you're building something ambitious and want an engineer who can own it end to end — and who already builds with AI, not around it — I'd love to hear from you.