Go Visualize

Marketing teams spend way too much time dragging pixels around design tools. Every social post, every ad, every flyer. Same grind, different day. And the moment you need something on-brand, you are digging through hex codes and font files again.

Go Visualize kills that workflow. Type what you want. AI generates production-ready, editable vector designs with your brand colors, typography, and layouts. Tweak if you want. Ship when ready.

I built a vector editor on top of Fabric.js. Multiple artboards in one workspace. Real-time manipulation of text layers, graphics, shapes, and grouped elements. The canvas handles simultaneous rendering across artboards without dropping frames.

I wired up a dual-agent AI system with Gemini 3 Flash Preview and Claude. You type a prompt. Both agents fire in parallel. One handles layout, the other handles visual styling. The results merge into a single design.

Parallelization is the key. Both agents running simultaneously inside the Next.js backend means generation takes seconds, not minutes.

The most technically fun part. Drop in a company URL. The platform scrapes the public website and extracts the entire brand identity. Colour swatches, typography hierarchy, logo positioning. The whole visual language loads directly into the canvas editor. No manual setup, no brand kit uploads. Just a URL.

Authentication runs on Better Auth with session persistence. Three billing tiers: usage-based, monthly, annual. All wired through Stripe’s Clover API. The twist is a credit ledger system that lets users buy ad-hoc token packs for heavy AI generation, separate from their subscription. Self-serve, no support tickets needed.

As Technical Lead, I defined the architecture. Prisma schema, component patterns, repo structure. Built the entire core system from February to early April 2026. Once it was stable and optimised, I handed off a production-ready codebase to the internal team. Shipped a complex multi-agent AI product in under two months.