Back to projects
Divorce Stoppers AI
Professional

Divorce Stoppers AI

Affaq Ahmed · AI Engineer · DevCache · AI / January 31, 2026

  • Next.js
  • Supabase
  • Claude API
  • OpenAI
  • Embeddings
  • PayPal
Live Demo

Overview

Divorce Stoppers AI is an AI-based relationship-coaching platform designed to improve communication and reduce divorce risk — helping people fix their texts and conversations using proven coaching methods, delivered through AI.

My Role & Contributions

I built the AI core of the product:

  • Complete RAG system — designed and built the full retrieval-augmented generation pipeline.
  • Vector database — implemented vector storage and search.
  • Embedding pipelines — built the pipelines that turn content into embeddings.
  • OCR integration — so users can upload screenshots of conversations to be read and analysed.
  • LLM integrations — Claude and OpenAI.
  • Payments — PayPal integration.

Key Features

  • AI relationship coaching grounded in a proven method via RAG.
  • Message analysis — upload texts (including screenshots via OCR) for advice.
  • Vector search over coaching knowledge.
  • Payments through PayPal.

Tech Stack & Architecture

  • App: Next.js with Supabase.
  • AI: Claude and OpenAI, with OpenAI embeddings and a vector database for retrieval.
  • Input: OCR for ingesting screenshots of conversations.
  • Payments: PayPal.

Challenges & Solutions

  • A full RAG pipeline — embeddings, vector storage, and retrieval so the coaching draws on the right material rather than generic advice.
  • OCR ingestion — letting users upload screenshots of their conversations and reliably extracting the text for analysis.

Timeline

December 2025 – January 2026 · 2 months

Links

Live at divorcestoppers.ai.