Cohult AI

Africa's AI adoption gap is not a talent problem. The continent produces world-class researchers, and the economic case for deployment is well-documented. The gap is infrastructure. Enterprise buyers operate on variable bandwidth, with legacy COBOL banking systems, data sovereignty requirements, and currency volatility that breaks pricing models built for stable economies. The pipeline from promising pilot to operational deployment is where value dies, and it has been dying for years. Generic AI marketplaces are not built for this context. They assume stable APIs, reliable connectivity, and buyers who can evaluate solutions without worrying about whether a model will fall over during load-shedding. CohultAI was built to fix that pipeline by creating a curated, high-conviction ecosystem that vets AI solutions specifically against African operational realities.

I built the entire platform from scratch including architecture, design, copy, and backend infrastructure. The site is split into two independently deployed units: a fully static Astro frontend with zero JS overhead for variable-connection environments, and a standalone Cloudflare Worker that handles all form submissions at the edge with no cold starts and no server to maintain. The Worker is hardened with CORS origin enforcement, body size guards before reading, honeypot bot detection, CRLF injection protection on email headers, a hard 10-second timeout race on Resend calls, and retry logic that distinguishes transient failures from permanent ones. Three separate institutional audience portals (enterprises, labs, and investors) each have their own UX flow and conversion logic. The Labs intake form includes a TRL field (Technology Readiness Level, the 1 to 9 scale used by NASA and serious R&D institutions) to filter for seriousness. Email confirmations are React Email components engineered for cross-client rendering with Playfair Display and a Georgia fallback for Gmail and Outlook, Outlook-compatible multi-column layout, and gradient strip fallbacks for clients that strip CSS. The Insights section contains original market intelligence research written from scratch as part of the product. A Cobe WebGL globe visualizes Co-HULT's geographic footprint across Lagos, Nairobi, Cape Town, and Accra.

• Live with real institutional clients across Africa

• Three-portal marketplace: enterprises, labs, and capital in one ecosystem

• Hardened edge backend with bot protection, injection guards, and retry-safe email pipeline

Technology Stack