Test your advertisement before the world does.
Ad-Versary predicts whether an advertisement will generate profit, engagement, or public backlash before it is launched.
Every year, businesses spend millions on advertisements that fail to connect with their audience.
Some campaigns are ignored.
Some fail to generate conversions.
Others unintentionally create public controversy, damaging brand reputation and wasting valuable marketing budgets.
Traditional market research methods such as surveys, focus groups, and manual reviews are often slow, expensive, and incapable of capturing real-world audience reactions at scale.
Ad-Versary solves this problem using AI.
Ad-Versary is an AI-powered advertising testing ground that simulates how different audience groups are likely to react to an advertisement before it goes live.
Users upload an advertisement in the form of:
- Images
- Videos
- Audio files
- Marketing scripts
- Social media creatives
- Product descriptions
The platform then creates an ecosystem of autonomous AI agents representing different audience personas, marketing experts, psychologists, cultural analysts, ethical reviewers, and adversarial critics.
These AI agents debate, challenge one another, and evaluate the advertisement using real Bangladeshi social, cultural, and behavioral intelligence.
Within 60 seconds, Ad-Versary generates a detailed business intelligence dashboard containing:
✅ Success Prediction Score
✅ Audience Engagement Forecast
✅ Brand Safety Analysis
✅ Cultural Sensitivity Assessment
✅ Potential Backlash Detection
✅ Actionable Campaign Improvements
Advertising success is difficult to predict before launch.
Businesses often face:
- Wasted marketing budgets
- Poor audience engagement
- Misaligned messaging
- Cultural misunderstandings
- Brand reputation damage
- Social media backlash
- Low conversion rates
The challenge becomes even greater in localized markets like Bangladesh, where cultural context, language, humor, trends, and social behaviors change rapidly.
Existing AI tools provide generic recommendations but fail to understand local audience dynamics.
Ad-Versary bridges this gap by creating a localized synthetic audience simulation system powered by GraphRAG and Multi-Agent AI.
Ad-Versary enables businesses to test advertisements inside a virtual AI-powered audience environment before spending money on deployment.
Instead of relying on a single AI model, the platform generates multiple autonomous agents that simulate realistic audience behavior.
The result is a highly contextual, explainable, and culturally aware evaluation process capable of identifying opportunities and risks before launch.
User Uploads Advertisement
│
▼
Frontend (Next.js)
│
▼
FastAPI Backend
│
▼
Multimodal AI Processing Layer
(Gemini Vision + Whisper + Claude)
│
▼
GraphRAG Intelligence Layer
(Neo4j + PGVector)
│
▼
Agentic Swarm Simulation
(LangGraph)
│
▼
Judge AI
│
▼
Analytics Dashboard
The user uploads:
- Video Advertisement
- Image Advertisement
- Audio Advertisement
- Campaign Script
- Marketing Brief
through the web interface.
The AI extracts:
- Visual Elements
- Speech Content
- Marketing Intent
- Emotional Tone
- Audience Assumptions
- Persuasion Techniques
- Brand Messaging
using advanced multimodal models.
The GraphRAG engine retrieves:
- Bangladeshi Social Trends
- Consumer Behaviors
- Platform Preferences
- Cultural Knowledge
- Emotional Patterns
- Audience Interests
from Neo4j and Vector Databases.
AI agents begin debating.
Examples include:
- Urban Teenagers
- University Students
- Rural Consumers
- Family-Oriented Buyers
- Luxury Buyers
- Marketing Expert
- Psychologist
- Cultural Analyst
- Ethical Reviewer
Agents specifically designed to attack the campaign and identify:
- Hidden Risks
- Offensive Messaging
- Brand Safety Issues
- Potential Viral Backlash
A final Judge Agent evaluates all arguments and generates a structured decision.
Users receive:
- Success Probability
- Engagement Forecast
- Brand Safety Score
- Cultural Risk Analysis
- Recommendation Report
within 60 seconds.
Responsible for understanding uploaded advertisements.
- Gemini Vision
- Whisper
- Claude
- Image Understanding
- Video Understanding
- Audio Transcription
- Emotion Detection
- Narrative Analysis
- Audience Inference
Provides contextual knowledge to AI agents.
- Neo4j
- PGVector
- LangChain
- LlamaIndex
- Social Trends
- Consumer Reviews
- Public Discussions
- Digital Communities
- Cultural Context
Orchestrated using LangGraph.
Simulate realistic customer reactions.
Evaluate campaign performance.
Analyze emotional influence.
Detect localized sensitivity issues.
Review compliance and fairness.
Actively search for vulnerabilities.
Unlike generic advertising tools, Ad-Versary is specifically designed for Bangladeshi audiences.
The platform understands:
- Bangla Language
- Local Humor
- Cultural Norms
- Religious Sensitivities
- Social Trends
- Consumer Behavior Patterns
This significantly improves prediction accuracy compared to globally trained models.
- Next.js
- TypeScript
- Tailwind CSS
- ShadCN UI
- Framer Motion
- FastAPI
- Python
- WebSockets
- Redis
- Gemini Vision
- Claude
- Whisper
- LangGraph
- Neo4j
- PGVector
- LangChain
- LlamaIndex
- Supabase Auth
- PostgreSQL
- Supabase
- Docker
- Vercel
- Railway / Render
Ad-Versary aims to:
- Reduce advertising failures
- Prevent PR disasters
- Improve campaign engagement
- Increase marketing ROI
- Enable data-driven creative decisions
- Localize campaigns effectively
- Advertisement Simulation Engine
- Bangladeshi GraphRAG
- Multi-Agent Debate System
- Real-Time Trend Tracking
- Cross-Platform Campaign Analysis
- Social Media Prediction Models
- South Asian Market Expansion
- Competitive Campaign Benchmarking
- Enterprise Intelligence Suite
Team Leader
Backend / Database / Scraper Engineer
Business Analyst / Data Scientist
Backend / Database / Scraper Engineer
UI/UX / Frontend Developer
Presentation / Communication Lead
Ad-Versary was developed by YSoB Team Infinity, a team under Young Sparks of Bangladesh, for BuildFest 2026.
Our mission is simple:
Transform advertising from guesswork into intelligence.