Outtie

Overview

We saw firsthand the problems publishers and creators dealt with finding ways to monetize their content naturally and in non-overt ways without the end user being bogged down by annoying ads, product pushes, or extraneous efforts to optimize SEO. So, we decided to develop a solution for this — which was the initial Outtie product — during a weekend hackathon.

Shortly after building our MVP and getting our first couple customers (one being a tech newsletter on beehiiv and the other being the substack newsletter by the dir. of research at Stonks), we were referred by one of them to the developer of a popular chrome extension with over 20k users. After landing this customer from the referral, we quickly realized that there was a massive potential to not only leverage our software with content publishers, but also with the underlying content that users of consumer apps interact with.

How it Works

1. Outtie scans the contents of a page (e.g., a Substack newsletter, TechCrunch article, social media post, etc.)
2. Our backend leverages AI to extract relevant keywords and links, and cross-reference those results with our database of 1000+ partner brands
3. It then converts these results to affiliated hyperlinks that generate commissions upon clicks and/or purchases.
4. Outtie then seamlessly updates the content of the page with these commissionable touch-points with no apparent changes to the end-user experience

Frontend + Backend Work

• Built Javascript Library / NPM package, HTML Script, and Google Tag
• Integrated OpenAI’s GPT-3 text-*davinci**003* model (now GPT-4) to extract relevant keywords and links from content
• Built and deployed the AWS infrastructure
  • EC2
  • RDS (PostgresSQL DB)
  • Lamda
  • Serverless
• DB models/architecture
• API design and implementation
• Integrating OpenAI’s Embeddings API with Pinecone for long-term memory & retrieval of keywords, ads/brands data to improve personalization over time