Instant Campaigns: Pioneering AI Social Campaign Generation

Hours saved per campaign

5-7

Engagement increase reported

400%+

Years in production

4

• React (Front end)
• Express (Web Server)
• Apollo GraphQL (API)
• PostgreSQL (Database)
• OpenAI
• Open Humans (API)
• JavaScript (App)

The Problem

Marketing teams run webinars — live online talks — to reach new customers. After each webinar, a team must turn it into days of social posts and emails. This work takes hours. It also needs a writer, a designer, and a manager to check each post before it goes live.

Contentware, a marketing technology startup, wanted to remove that manual work. The founder needed a platform that could read a webinar's landing page, write a full multi-day campaign from it, and let a human check and approve every piece before it went out.

What I Built

I was the lead architect and full-stack developer for Contentware from 2020 to 2024. I designed the system from the ground up. I built most of it myself, with support from other engineers, a data scientist, and a cloud infrastructure specialist.

Application layer. I built the React admin app for the Contentware team and the React client app for paying customers. Both shared a component library and a design system. Both apps ran on a modular GraphQL API, built with Apollo GraphQL and Node.js. That API handled campaign creation, content review, approval, publishing, and billing. A dynamic campaign calendar showed the full timeline with previews that matched how each post would look on its target platform.

Content generation. The platform pulled text and images from a customer's webinar landing page. It sent that content to the OpenAI API to draft a multi-day campaign: a sequence of social posts and marketing emails. I built a custom templating language that let customers specify exactly what each post should say, so the AI drafted from a precise brief instead of a vague prompt. Each customer could also set campaign rules — tone, post count, platform mix, posting cadence — which shaped how the draft came out.

Human review. No AI draft published on its own. Every generated post and email went through a review screen. There, a customer or a Contentware team member could edit the text, approve it, or send it back for another draft. This human-in-the-loop step was central to the product. Customers trusted the output because they had the final say.

Data and integrations. I designed the PostgreSQL schema and API structure to handle campaign, account, billing, and integration data at speed. A customer could open a campaign and see every post's status without a slow page load. I integrated the platform with Facebook, Instagram, Twitter/X, and LinkedIn for publishing; Mailchimp, Constant Contact, HubSpot, Informz, Omeda, and ActiveCampaign for email and marketing automation; Hootsuite for social scheduling; and Chargebee for billing and subscriptions.

Infrastructure. I set up and maintained separate development, staging, and production environments on AWS. CI/CD pipelines made deploys safe and repeatable for a live product with paying customers.

My Role on the Team

I owned the architecture and wrote most of the application code. I also coordinated day-to-day work with two other engineers. A data scientist tuned the content generation prompts and rules. A cloud infrastructure specialist managed AWS. I worked directly with the founder to turn product goals into a technical plan, and I made the final call on architecture and delivery tradeoffs.

Impact

Contentware moved campaign creation from a manual, hours-long task into a deployed SaaS workflow with built-in review and approval. The platform ran in production for four years and served paying customers.

One customer reported a 400%+ increase in engagement for their webinars after adopting Contentware-generated campaigns.