RankBee.ai
AI Visibility & Brand Optimization Platform
Tech stack
Architecture highlights
- Queue-based ingestion (Bull) for fault-tolerant LLM querying
- PostgreSQL for relational analytics + citation history
- Clerk auth, Sentry + Datadog observability, Intercom support
The problem
Brands had no way to track how they appear across AI answer engines (ChatGPT, Claude, Gemini, Perplexity) as users shifted from Google to AI search.
What I built
A full-stack platform (Next.js + Nest.js) — built for a VC-backed startup founded by an ex-Amazon Head of SEO — that runs 100K+ daily AI queries and surfaces brand visibility, citations, and competitive rankings in a real-time dashboard.
Key features
- Automated brand & competitor visibility tracking across 4 major LLMs
- Real-time analytics dashboard with citation & sentiment insights
- Marketing site + reputation alerting for enterprise teams
- Scalable job queue processing 100K+ AI queries per day
The hardest challenge
Coordinating rate limits, cost, and freshness across four different LLM APIs at 100K+ queries/day without blowing the budget — solved with a Bull queue, aggressive caching, and per-provider throttling.
The full story
Problem
As buyers shifted from Google to AI assistants, brands lost visibility into how — or whether — they were being recommended. There was no analytics layer for the new AI search surface.
Research
I benchmarked each LLM's API for latency, cost, and citation behaviour, then modelled query volume and budget to find a sustainable querying strategy at scale.
Architecture
Next.js frontend + Nest.js API, a Bull queue for fault-tolerant LLM querying, PostgreSQL for analytics and citation history, and Clerk / Sentry / Datadog / Intercom for auth, observability and support.
Implementation
I built the ingestion workers, per-provider rate limiting and caching, the analytics dashboard with real-time visualisations, and the marketing site — coordinating four LLM providers behind one consistent data model.
Result
The platform reliably processes 100K+ AI queries per day and gives enterprise teams a real-time view of brand visibility, citations and competitive rankings.
Lessons learned
Queue-first design and aggressive caching were the difference between a viable unit economics story and a runaway API bill. Observability from day one paid for itself many times over.
Outcome
- 100K+ daily AI queries processed reliably
- Enterprise-ready analytics adopted by early customers
- Full observability via Sentry, Datadog & Intercom