WildAware — A Ranger in Every Pocket

Picture the front line of wildlife trafficking. It isn't a jungle checkpoint or a customs hall. It's a market stall in a tourist town — a carved trinket in a traveler's hand, and a single unanswerable question: is this bone, resin, or elephant ivory? The multi-billion-dollar illicit wildlife trade moves through exactly that moment, millions of times a year, because the person holding the object has no way to know and nowhere to report.

Now imagine that traveler lifts their phone, photographs the trinket, and thirty seconds later holds a risk assessment, the suspected species, the CITES legal context, and an ethical alternative — with a one-tap path to report it to a conservation foundation whose triage queue lights up in real time.

That's WildAware. The front line didn't move. The ranger did — into every pocket.

The Client

Green Dolphin Project Global is a wildlife conservation nonprofit whose community — travelers, volunteers, and contacts through The Explorers Club network — kept colliding with the moment of purchase and coming away empty-handed. Suspicious sightings scattered across email, WhatsApp, and word of mouth. The foundation had no intake queue, no triage, no status tracking, and no visibility into who in their community was active. And the conventional fix — a custom app with auth, database, AI vision, storage, and an admin console — is a three-to-six-month, six-figure agency engagement. The foundation runs no servers and employs no engineers. Under the conventional model, this project doesn't get built. Full stop.

The Build: The Whole Loop, Not Just an App

Rocket Now delivered WildAware as a mobile-first, installable Progressive Web App covering the complete loop — identify → learn → report → triage → follow up — because a scanner without a reporting path is a curiosity, and a reporting form without triage is an inbox. The loop is the product.

Identify. The AI Product Scanner takes a camera capture or gallery photo and returns a tiered advisory — low, medium, or high risk — with suspected species, CITES and Lacey Act legal context, and ethical alternatives. The vision model runs server-side through a managed gateway inside a single edge function, so no credential ever reaches the client, and the output is constrained to a strict JSON schema parsed defensively — the interface never renders free-form model prose. When the model is rate-limited or credits run dry, then the user receives a safe, non-alarming fallback instead of a dead end. And deliberately, everywhere in the product: the result is an advisory with legal context, never a verdict — and the guidance says observe and photograph only where safe and legal, never confront.

Report. Every scan persists to history with its photo. When a scan comes back medium or high risk, then one tap escalates it into a pre-filled incident report — the AI findings and the same stored photo carried through router state, so the report references the exact artifact the model analyzed. Four manual steps collapsed into one, and the reporter walks away with a reference number.

Triage. The admin portal turns the foundation from an inbox into an operation: new reports appear instantly via realtime subscriptions with toast notifications, carry status through resolution, and every user has a detail page showing their complete scan and report history. A news system — drafts, cover images, feed, and article pages — lets the foundation publish to its community without a webmaster.

Underneath it all, the same security spine we build into fintech platforms: row-level security on every table with 25+ policies, owner-scoped storage, and roles held in a dedicated table behind a security-definer check — because a conservation nonprofit's reporters deserve the same protection as a bank's depositors.

One decision made the economics work before a line was written: PWA over native. When app-store review cycles, dual codebases, and developer-account fees serve the platform more than the mission, then skip them — WildAware installs to the home screen, uses the camera, and ships an update the moment it's ready.

The Delivery: Weeks, on a Nonprofit Budget

The engagement ran as a conversational build — the client's specs, brand documents, and mockups went in; deployed software came out, often the same session. The first working MVP was live the same day the brief arrived, with the full feature set iterated over roughly four months of part-time cadence. Scope discovery happened inside the engagement: scan history, one-tap escalation, the news system, and admin user profiles all emerged from client questions mid-build and shipped the same day they were asked for. Under a fixed-scope contract, each of those is a change order; here, each was an afternoon. The foundation estimates the conventional equivalent at $80k–$150k over three to six months — and the truer number is that under the conventional model, the project simply wouldn't exist.

The app is live in production today at wildaware.app, in early access with the foundation's core team and Explorers Club contacts, ahead of a structured 50–100 volunteer pilot across high-incidence regions and a public launch to follow — with the pilot's success defined the right way: labeled scanner-accuracy review, triage response times under 48 hours, and a retention bar the badge system and news cadence are built to clear.

The Results

A production AI vision application, live on a custom domain, operated by a nonprofit with zero servers and zero engineering staff
The complete identify-to-triage loop delivered: tiered AI risk advisories, persisted scan history, one-tap escalation, referenced incident reports, and a realtime admin queue with status tracking
Four manual triage steps reduced to one for every flagged item — photo, description, species, and legal context auto-carried from scan to report
Security at fintech grade: RLS on 100% of user data tables, isolated role storage, server-side-only AI credentials with graceful degradation
An estimated $80k–$150k conventional build cost avoided — and a project that exists where the conventional model says it can't

"[Quote from Green Dolphin Project Global — pending sign-off.]" — [Name, Title], Green Dolphin Project Global

Why It Worked: One Right Step, at the Moment of Purchase

The Rocket Now principle, one more time: a single step in the right direction is worth exponentially more than many steps in the wrong one. The wrong steps here were the familiar ones — an awareness website, a PDF field guide, a native-app odyssey through two app stores. The one right step was aiming the entire build at the moment of purchase and closing the loop around it: the scan that becomes a report that becomes a triaged case, in one motion, in the field. Every hour of the build served that single moment. When the mission has one decisive moment, then the product should have one decisive path — and everything else is decoration.

Have a mission the conventional software model prices out of existence? That's exactly the kind of math we like to break. → RocketNow.com

Editor's note — before publishing: