Pipelines to People (P2P) started as a simple data pipeline—a way to pull creator data from YouTube and rank channels based on subscriber count, niche, and engagement. But as I dug deeper into the world of creator-brand collaborations, I realized raw metrics weren’t enough. Brands didn’t want more data—they wanted qualified, brand-aligned creators they could trust.
What began as a scraping tool evolved into a full-blown system for lead discovery, enrichment, scoring, segmentation, and outreach—automated from end to end. Today, P2P is a modular architecture that uses AI agents, enrichment APIs, and automation workflows to connect brands with the right creators, at the right time, with the right message.
At the beginning, I built a simple Python-based ETL pipeline that pulled data from the YouTube API. It fetched channel info—like titles, subscriber counts, and video engagement—and stored it in a structured format. I used EDA to surface patterns and spot creators that stood out based on performance metrics.
Once the data pipeline was in place, I started manually reviewing creators to find collaboration-ready leads. But the manual vetting didn’t scale. I needed a better way to qualify creators.
To remove the bottleneck, I integrated the pipeline with n8n. Every time the ETL pipeline finished, it would POST top creator leads to a webhook. n8n took over from there:
At this point, I had a working system that not only sourced creators but also qualified and contacted them, with minimal input required.
As the system matured, I added modular AI agents: