influenca

Natural language search to find and book local influencers. Think OpenTable for influencer marketing.

Comment

Inspiration

Local businesses — cafes, restaurants, pop-ups — want to work with influencers but the tools that exist cost $2,000+/month and are built for Fortune 500 brands. Business owners are stuck manually scrolling Instagram and DMing creators. We wanted to build something where you just type what you need and get results.

What it does

Type a query like "10 food micro-influencers near Japantown for a ramen shop opening" and get a ranked list of real creators with match scores, engagement rates, and contact info. Search mode with filters or chat mode for iterative refinement.

How we built it

The core problem was building an influencer database without scraping. Our pipeline:

  1. You.com Search + Contents API — mined blog posts, influencer lists, and local press to discover ~2,000+ SF creator profiles from the public web (~$15 in API credits)
  2. Claude API — classifies each profile (niche, quality score, neighborhoods) and parses natural language search queries into structured database filters via tool-use
  3. PostgreSQL + pgvector on Google Cloud SQL, Next.js frontend on Cloud Run

Total infrastructure cost: under $20/month.

Challenges

  • Instagram handles appear in dozens of formats across articles — required multi-pattern extraction with false-positive filtering
  • Same influencers appear on 10+ lists — aggressive deduplication before spending enrichment credits
  • $100 API budget meant prioritizing which profiles to enrich deeply vs. deferring to on-demand lookup
  • Mapping informal queries like "coffee content around Hayes Valley or the Marina" to structured SQL reliably

What we learned

You don't need 200M generic profiles to be useful. 2,000 well-profiled, locally-relevant creators in one city is more valuable to a small business than any enterprise database. The indexed web + LLMs can bootstrap a vertical dataset surprisingly cheaply.

What's next

Creator self-registration via a free content tool (script generator), post-campaign rating system to build proprietary performance data, and expansion to Austin and Nashville.

Built With

Share this project:

Updates