Case study
GluonAI
AI-powered email marketing platform with campaign management, an editor, and generation of on-brand content + HTML designs.
Overview
GluonAI is a full-stack email marketing platform that helps teams generate email campaigns faster. It pairs a Next.js web app (campaigns, editor, assets, subscriptions) with a FastAPI service that performs market-research style analysis and generates structured email content and responsive HTML designs.
Problem
Email campaigns are repetitive and time-consuming: marketers need subject lines, intros, body copy, CTAs, and a layout — all consistent with the brand. Manual workflows slow iteration and create inconsistency across campaigns.
Solution
I built the product end-to-end: a Next.js application for campaign lifecycle and editing, plus a FastAPI backend that generates multiple variations of email content and produces HTML email designs (including image-aware layout decisions).
Architecture
- Next.js app → collects brand tone + product information + assets (Supabase Auth/DB/Storage).
- AI text endpoint → returns structured variants (subject, intro, body, CTAs).
- AI design endpoint → generates mobile-friendly HTML template and places images based on analysis.
- Stripe tiers + webhooks → enforce limits and subscription access.
Tech stack
Key engineering decisions
- • Generated structured content blocks (subject/intro/body/CTA) to keep outputs predictable and editable.
- • Separated 'text generation' from 'design generation' to reduce coupling and simplify debugging.
- • Tier-based limits aligned compute cost with pricing, enforced server-side via Stripe subscription state.
- • Used image analysis to improve layout choices and avoid broken email composition.
Results
- • 70% reduction in email template creation time.
- • Generated multiple subject/intro/body/CTA variants per request for faster iteration.
- • Delivered responsive HTML designs with image-aware layout generation.
Links
What I’d improve next
- • Add an evaluation harness (quality + brand-consistency scoring) and regression tests for prompts.
- • Introduce caching + streaming for better perceived speed and lower repeat costs.
- • Add observability: tracing per generation, error budgets, and cost dashboards.