| The Reality: They didn't. But the tech is real. I’ve been building an AI SDR platform and I wanted to share the stack with the HN crowd. The Project: Babuger
Babuger automates the entire outbound/inbound lifecycle. It trains on your best rep's scripts to qualify leads, handle objections, and book meetings 24/7. The Problem: Traditional SDR teams are expensive ($150k/yr), have high turnover, and ignore "dead" leads. The Solution: One human orchestrator managing 20+ specialized AI agents. The Result: 90% task automation and 70% response rates on neglected pipelines. The Tech Stack
I kept it modern and modular to handle complex multi-step reasoning: Agent Orchestration: LangGraph. This was crucial for handling non-linear conversation flows (loops, conditional routing, and state management) that standard DAGs can't touch. LLM Framework: LangChain. Used for prompt templating, output parsing, and integrating various toolsets (Gmail/Cal.com/HubSpot). Frontend: Next.js. Managed the dashboard, live email thread previews, and real-time pipeline analytics. Why I’m Posting
I’m looking for the "HN stress test." Is the agentic approach with LangGraph the right move for scaling to 10k+ interactions/mo, or should I be looking at a more custom state machine? Check it out: Babuger.com |