KĀRYO Lead Intelligence Agent
An automated lead generation and qualification pipeline built to support KĀRYO's door-to-door sales efforts in Bangalore.
The Problem
Manually researching local businesses, checking if they have a website, and drafting personalized outreach is incredibly time-consuming and scales poorly.
The Solution
A multi-agent system using CrewAI. Agent 1 scrapes local directories. Agent 2 qualifies leads based on digital footprint. Agent 3 drafts personalized outreach. Agent 4 reviews and formats the output.
Architecture
Python scripts orchestrating CrewAI with Groq as the LLM engine for cost-effective, rapid reasoning. Outputs are structured using Pydantic and exported to .docx.
Engineering Challenges
Preventing hallucinations in the qualification agent. Enforced strict JSON schema outputs using Pydantic to ensure reliable data passing between agents.
Results & Impact
Significantly accelerated KĀRYO's prospecting phase, generating dozens of highly qualified leads per run.
Lessons Learned
Learned the complexities of multi-agent orchestration and the importance of structured outputs in LLM chains.