Aurora
Market research, Consulting
An AI market-research platform with competitor analysis and a LangGraph-based TAM/SAM calculator, turning a founder's research process into self-serve tools.
AI Automation & Agents
Cost-aware multi-engine AI search for GEO
A custom search and synthesis engine that combines several search APIs, live scraping and routed LLMs to deliver GEO insights, such as how a brand appears in AI answers, at a lower cost per query.
- Google, Bing and Brave results merged for each query
- Multi-engine
- OpenAI, Claude, Gemini and DeepSeek, routed per task
- Multi-model
- Answers synthesised with their sources
- Cited
The starting point
The challenge
Tracking how brands appear in AI search means running many searches and model calls for every query. Paying premium rates for each query and analysis run adds up fast, and the client wanted the same quality of insight at a lower cost.
The solution
What we built
Built a FastAPI service that analyses each query and checks a Redis semantic cache first, so repeated and similar queries return instantly.
Routed searches through SerpAPI across Google, Bing and Brave, then merged and de-duplicated the results.
Scraped the top sources in real time with Crawl4AI, including pages that need a headless browser to render.
Cleaned, chunked and ranked the content by relevance, keeping source attribution for citations.
Routed each task to a suitable model across OpenAI, Claude, Gemini and DeepSeek through OpenRouter, with fallbacks and retries.
Synthesised answers across documents with cited sources, and reported GEO measures such as brand visibility, citation coverage and source ranking.
Architecture
How the system works
Engineering
How it was engineered
Controlled cost at every layer: a semantic cache, task-based model selection and token-aware routing.
Kept the model layer LLM-agnostic, so models can be swapped without touching the pipeline.
Scope
What was delivered
01
Query analysis and semantic cache
02
Multi-engine search routing
03
Real-time scraping with Crawl4AI
04
Content ranking with attribution
05
Multi-LLM routing with cost control
06
Cited synthesis and GEO analytics
Technology
Stack
FastAPILangChainSerpAPICrawl4AIRedisOpenRouterOpenAIClaudeGeminiDeepSeek
More work
Related case studies
Narratize
Product innovation, R&D
Primary AI development partner for Narratize's earliest generative features: document ingestion, technical briefs, research reports and scientific analysis.
Booosted
SaaS, Customer support
The retrieval engine behind a Finnish AI-chatbot product: businesses add their website or PDFs and get a grounded chatbot they can embed on their site.
Need something similar?
We can scope and build a system with comparable capabilities, shaped around your own workflows, data and integrations.