AI Content Pipeline for Quackmate

Quackmate was an automated social media content platform for small businesses. I co-founded the project after seeing how much of Vooyai's social media workflow could be automated with generative AI.

Problem: Small businesses often need consistent social media output but lack the time, tooling, or creative workflow to generate posts, visuals, scheduling plans, and analytics in one place.

Technical scope: I built the automated generation workflow, scheduling and analytics features, media storage pipeline, and prompt-engineering layer for content and image generation. This included designing prompts and application logic to produce predictable outputs at a time when LLMs and image models were less reliable and structured-output support was still limited.

Architecture: The application used Next.js, Supabase, Cloudflare R2, DALL-E 3, and Vercel.

Implementation: The product interface and serverless workflows were built with Next.js. Supabase handled authentication, persistence, and product data, while DALL-E 3 powered generative media creation and Cloudflare R2 provided low-latency object storage for generated assets.

The initial plan was to use a backend closer to Vooyai's AWS setup, but the project shifted toward Vercel-hosted serverless functions to reduce operational cost and complexity.

Outcome: The project reached a working MVP with automated content creation, media generation, scheduling, analytics, and asset storage. It also gave me deeper experience in building product workflows around generative AI systems rather than isolated model calls.