BioJalisco — Biodiversity Atlas Pitch Site + Species Identifier

Overview

BioJalisco is a proposed citizen-science biodiversity monitoring platform for western Mexico. Before building the platform, the project needed a persuasion artifact — a site that communicates the vision with enough emotional and intellectual weight to win over academic partners, conservation organizations, and potential funders.

This scrollytelling site is that artifact. Built as a single self-contained HTML file, it uses a cinematic 5-act story arc (Hook → Problem → Vision → Evidence → Ask) with scroll-triggered animations, parallax photography, ambient forest audio, and bilingual content to make the case for BioJalisco.

Every image is embedded as a base64 data URI, making the site fully portable — it works offline, deploys anywhere, and has zero external dependencies beyond Google Fonts.

The Challenge

• The reach problem: Verónica Rosas-Espinoza has deep expertise in Jalisco's biodiversity, but traditional ecology covers only ~5% of the territory systematically. The pitch needed to make this gap visceral, not abstract.
• Dual audience: The site must work for Spanish-speaking Mexican academics and English-speaking international partners — not as separate pages, but as a seamless bilingual experience.
• Persuasion, not information: This isn't a documentation site. It's a narrative designed to move people from awareness to action. Every section must earn its scroll.
• Zero infrastructure: The site needed to deploy instantly with no build step, no CMS, no backend — just drop a file on Vercel.

The Solution

Cinematic narrative structure: A 5-act story arc modeled on screenwriting principles. The hero section poses a provocative question, the problem section makes the data crisis tangible with animated counters, the vision section introduces BioJalisco as the answer, the evidence section builds credibility with a roadmap and proof points (iNaturalist precedent, book trilogy, team credentials), and the closing ask creates urgency.

Bilingual architecture: A data-lang attribute on the body element combined with .lang-es / .lang-en CSS classes enables instant language switching with zero page reload. All content exists in both languages in the DOM — CSS toggles visibility.

Self-contained deployment: All 9 images converted from PNG to WebP (88-95% size reduction), then embedded as base64 data URIs. The entire site is a single 1.8MB HTML file with no external asset dependencies.

Procedural audio: The ambient forest soundscape is generated in real-time using Web Audio API — layered noise filters for wind, oscillators for bird chirps, and a low drone for depth. No audio files needed.

Technical Highlights

• Base64 image pipeline: Python/Pillow batch conversion from PNG to optimized WebP, then base64 encoding — reduced total image payload from 16.8MB to 1.3MB (92% reduction)
• Intersection Observer scroll system: Multiple reveal classes (translate-Y, translate-X, scale) with staggered delays create a choreographed scroll experience without a single animation library
• Web Audio API synthesis: Three-layer procedural soundscape (filtered noise for wind, frequency-modulated oscillators for bird chirps, sine drone for depth) with fade-in/fade-out on toggle
• CSS-only parallax: background-attachment: fixed on the vision section creates depth without JavaScript scroll listeners
• Animated counters: Eased counter animations triggered by IntersectionObserver with cubic easing for natural deceleration
• AI Species Identification: GPT-4o vision API with structured JSON output — confidence percentage, full taxonomy (Kingdom→Species), ecology (habitat, diet, size, lifespan, behavior), geographic range with Jalisco/Mexico/invasive badges, IUCN conservation status, and similar species comparison. Handles wildlife, insects, plants, and domestic animals (breed identification with temperament and origin history).
• Tabbed results UI: Six-tab interface (Overview, Taxonomy, Ecology, Range, Conservation, Similar) with circular confidence gauge, taxonomy tree visualization, and scan history persisted in localStorage
• Serverless architecture: Same Python codebase runs as Flask dev server locally and as a Vercel serverless function in production

Results

For the Stakeholder Audience:

• Complete narrative pitch accessible in both Spanish and English
• Credibility elements (PhD photo, book trilogy, fieldwork photography) embedded directly
• Single URL to share — no login, no app install, works on any device
• Interactive Species Identifier demonstrates real AI capability — stakeholders can try it immediately

Technical Demonstration:

• Proves that a high-impact persuasion site doesn't need React, a build system, or a CMS
• Shows restraint in technology choices — vanilla JS for a project that genuinely doesn't need frameworks
• AI integration via serverless functions keeps the architecture simple while delivering real utility
• Image optimization pipeline that could be reused across CushLabs projects