YOGI TIMES needed to expand their global database of yoga teachers but faced a growth ceiling. Manual discovery was too slow to deduplicate against their existing 150k contact records, and identifying high-quality leads required hours of deep-dive research into company reviews and business models.
2. The Solution: A Multi-LLM Orchestration Pipeline
I engineered an end-to-end system using Airtable and Make. While originally prototyped for a specific client, I have refined this into a sector-agnostic architecture that can be redeployed for any niche.
Dynamic Search Queue: A custom Airtable search query builder that rotates through search variables to prevent stale results.
Mass Discovery (Perplexity AI): Uses the Sonar model for high-volume web retrieval, extracting dozens of raw leads per cycle.
Lead Qualification Detective (Gemini Pro): A secondary AI layer that ruthlessly filters leads by business model and operational status.
Structured Enrollment (Gemini Flash): Converts raw research into clean JSON data for the master database.
3. Impact & Scalability
Adjustable Frequency: The system can be throttled to match any outreach capacity.
100% Automated Deduplication: Through extensive use of Make's Data Stores and an external email list, no lead is ever processed twice, protecting the integrity of the 150k+ record database.
Universal Utility: The query builder is variables-based; swapping "Yoga" for "SaaS" or "Construction" allows the system to find any target persona globally.
Locations can be toggled on or off
Create a database of dynamic templates for search queries
Change the niche or persona to create fresh search queries
Create a search queue that feeds the Make scenario indefinitely