Clearfraim
The AI team a 10-person company can't afford.
AI-native data integration and BI for small teams. Connects Shopify, Meta Ads, Stripe, Notion, and Sheets — agents answer business questions with source attribution.
A 10-person company can't afford a data team — but it still needs answers from its data every day. The sources are scattered (Shopify, Stripe, Meta, Notion, Sheets), the questions are specific, and a generic LLM has no idea what "top SKU last quarter" means in your business. The gap is the canonical layer underneath, not the model on top.
Start with the data layer. Every connector lands in the same canonical schema, owned per-tenant, never trained on. Put a RAG agent on top that always cites its sources and is allowed to refuse. Wrap it in an eval harness that catches drift before customers see it. The model is the thin part; the stitch underneath is the moat.
- 01Shopify, Stripe, Meta Ads, Notion, and Sheets in one canonical schema
- 02Source-attribution on every answer — refuses when it doesn't know
- 03Eval harness gates every prompt or model change before deploy
- 04Multi-tenant, per-tenant data isolation, no model training on customer data
- 05Drift watcher reruns the eval set hourly and pages on regression
How this kind of thing gets built.
Most of the thinking behind builds like this one — the patterns, the trade-offs, what holds up — lives in the writing.
The rest of
the body of work.
One flagship live, more in build. Every product runs the same standardized stitch underneath.
