How I Built an AI Discovery Platform From Scratch
Back in April, I finally launched ZerenAI, a project I had been envisioning for months. My goal was to create a platform that helps small and medium-sized businesses, as well as developers, discover the right AI tools for their unique use cases. With so many new products flooding the market, finding trustworthy information is harder than ever. ZerenAI aims to make that process simple, actionable, and transparent. But taking an idea from a whiteboard sketch to a live AI discovery platform required careful technical and branding decisions.
I chose Next.js 15 as the foundation for ZerenAI, mainly because of its strong SEO features. Server-side rendering (SSR) and dynamic meta tags were critical, if you’re building an AI app with Next.js, you need your content to be indexable and fast. Migrating from my prototype React + Express setup wasn’t easy. I had to refactor mixed JavaScript and TypeScript components, solve hydration mismatch errors, and rethink routing. Still, the result was worth it: ZerenAI now has the scalability, performance, and SEO-friendly foundation needed for long-term growth.
For content management, I integrated Sanity CMS to power product pages, blog posts, and structured data about AI tools. I designed schemas that included meta descriptions, categories, and nested builder arrays so the platform could scale with new content types. Automating data ingestion became a project in itself. I wrote crawlers that pulled product information, converted it into NDJSON, and imported it into Sanity datasets. More importantly, I set up recurring jobs to refresh pricing and review, because an AI tool directory is only as good as its accuracy and freshness.
On the intelligence side, I tested both OpenAI endpoints and AWS Bedrock Agents. Bedrock gave me better options for knowledge base ingestion and recommendations, though parsing event streams required custom debugging. This let me build smarter filters and recommendation logic into ZerenAI, so users can search by category, explore tool comparisons, and land on SEO-optimized product pages. The result is a business-focused AI recommendation engine that helps visitors cut through noise and find tools that actually solve their problems.
The biggest lesson from building ZerenAI is that launching a modern platform means wearing multiple hats: engineer, product manager, marketer, and storyteller. Writing code was just one part. I also designed the front-end experience, optimized for accessibility, and created branded hero images that built trust. If you’re looking to build an AI app today, you can’t separate technical rigor from marketing clarity. The projects that stand out are the ones that blend both.
Ultimately, launching ZerenAI in April wasn’t about chasing perfection. It was about creating a solid, SEO-friendly foundation for future growth. Every decision, from choosing Next.js to designing Sanity schemas, shaped the platform into something usable and trustworthy. And as the AI ecosystem continues to evolve, my goal is that ZerenAI will be a resource not just for AI enthusiasts but also for businesses who need reliable guidance on which tools to adopt.