AI-powered threat intelligence API to detect scams, phishing, and fraudulent messages

I participated in the India AI Impact Buildathon at the AI Impact Summit 2026, alongside 40,000+ innovators across India.
During the hackathon, I built Agentic Honey-Pot — an AI-powered threat intelligence API to detect scams, phishing, and fraudulent messages.
Key Features
🔍 Multi-Agent Orchestration
7 specialized AI agents analyze suspicious messages.
🤖 Scam Classification
Using DistilBERT to classify messages into 10+ scam types (Phishing, Advance Fee, Tech Support, etc.).
🎯 Threat Scoring
Generates a 0–100 threat score with severity levels from Low → Critical.
🕵️ Intent Inference
Explains the attacker’s likely motivation in plain language.
🔐 Tech Stack
FastAPI • PostgreSQL • Hugging Face Transformers
Scam communications are increasing in volume and sophistication. Building tools to detect and understand these threats is critical.