- Year
- 2026
- Role
- Full Stack Developer & Product Design
- Status
- Shipped
01
The Problem
Small sales teams drown in manual data entry. Deals slip because follow-ups depend on memory, and existing CRMs treat AI as a bolt-on chatbot instead of a workflow layer.
02
The Solution
A MERN-based CRM where every contact interaction is embedded and indexed. An LLM pipeline summarizes call notes, scores deal health, and drafts context-aware follow-up emails — all reviewable in a single keyboard-driven inbox.
03
The Result
Pilot teams cut admin time by ~40% and doubled follow-up consistency. Sub-100ms interaction latency across the app with optimistic updates and a fully typed API.
Engineering highlights
- RAG pipeline over deal history with pgvector-style embeddings on MongoDB Atlas Vector Search
- Realtime deal board with Socket.IO and optimistic drag-and-drop
- Role-based access control and full audit trail
- CI/CD with preview deployments on every pull request
Stack
- Next.js
- TypeScript
- Node.js
- Express
- MongoDB Atlas
- OpenAI API
- Socket.IO
- Tailwind CSS
