AI Concept Testing with Cited Evidence
The AI layer of two industry intelligence dashboards (food and beverage, and real estate) for brand teams in the Gulf: concept testing, a tool-calling assistant, a trend pipeline, and a knowledge base, all grounded in social posts and client research.
Project Overview
The AI layer of two industry intelligence dashboards (food and beverage, and real estate) for brand teams in the Gulf: concept testing, a tool-calling assistant, a trend pipeline, and a knowledge base, all grounded in social posts and client research.
The Problem
Brand teams kept asking the same questions: what is trending in my category, how do my brand and competitors show up, and will this new idea land? The answers sat in thousands of Arabic and English posts and images, and brand teams would not act on a black-box score.
The Solution
FastAPI microservices over PostgreSQL with pgvector and OpenAI, behind a Next.js dashboard, where every AI answer carries its evidence.
Use Cases
Concept testing
Six-factor scores with explanations and citations to real posts and reports.
Explainable confidence
Confidence comes from evidence volume, relevance, recency, and spread, and is explained in plain language.
Grounded assistant
Answers come from tool calls into the data, not from the model's memory.
Trend pipeline
Posts become named, described trends, with movers and period comparisons.
Client knowledge base
Uploaded research PDFs are chunked, embedded, and cited next to social evidence.
AI cost tracking
Token use and cost recorded per answer and per test.
Project Screenshots
What I Delivered
- Concept testing with citations and a confidence score
- Tool-calling assistant grounded in the data
- pgvector trend pipeline tuned for the full dataset
- PDF knowledge base (RAG)
- Per-call AI cost tracking and a cost report
Ready to build something useful?
Whether you need a freelance build, a full-time hire, or a short intro call, pick the path that fits and I will respond within 24 hours.