The Next-Generation SaaS Platform for High-ROI Creator Matching
Overview • Features • Architecture • Demo • Setup
InfluenceIQ is an enterprise-grade AI analytics platform engineered to solve the multi-billion dollar problem of influencer misclassification and audience bot fraud.
In the modern creator economy, brands struggle with "vanity metrics"—high follower counts that don't translate to actual sales. InfluenceIQ solves this by providing a comprehensive suite of tools for:
- Discovery: Finding the perfect niche creator from a verified pool of 400+ influencers.
- Vetting: Real-time engagement analysis and automated fraud risk detection.
- Outreach: AI-generated negotiation scripts tailored to each creator's performance.
- Management: End-to-end campaign structuring and offer tracking.
By utilizing dense vector embeddings and heuristic machine learning pipelines, InfluenceIQ intelligently matches brands with the exact right creators. It bypasses shallow keyword searching to understand the deep semantic context of a campaign, executing real-time fraud analysis via live YouTube API telemetry.
Our Mission: Replace days of manual influencer outreach and guesswork with milliseconds of data-driven, mathematically proven AI curation.
The precision of InfluenceIQ is driven by several state-of-the-art models and custom algorithmic pipelines:
- Purpose: Converts complex creator profiles and brand requirements into 384-dimensional dense vectors.
- Why: Traditional keyword search fails if a brand says "Streetwear" and a creator uses "Urban Style". S-BERT understands these are semantically identical, ensuring high-quality matches even with different terminology.
- Fallback: For low-resource environments, the system includes a
TF-IDF(Term Frequency-Inverse Document Frequency) fallback to maintain basic functionality.
- Purpose: Measures the "angle" between the brand requirement vector and all 400+ influencer vectors in the database.
- Logic: This provides a mathematical "Suitability Score" from 0-100%, ranking creators based on how closely their content matches the brand's vision.
- Purpose: Analyzes the ratio of followers to engagement and likes to comments.
- Logic: It flags influencers with "Low", "Medium", or "High" fraud risk based on statistical deviations from platform norms, protecting brands from investing in bot-inflated audiences.
- Purpose: Fetches live subscriber counts, views, and engagement metrics directly from the source.
- Resilience Logic: Implements a fail-safe mechanism that detects quota limits (HTTP 403). When triggered, the system gracefully falls back to cached/pre-computed ML scores from the local dataset, ensuring 100% platform uptime even during API downtime.
- Impact: Ensures that the data is never stale when available, while maintaining complete discovery functionality via the local ML engine as a fallback.
- Purpose: Enables granular location-based filtering for targeted regional campaigns.
- Logic: Dynamically populates regional states (e.g., for India) and integrates them into the cosine similarity weighting engine, allowing brands to find hyper-local creators in specific territories like Maharashtra, Delhi, or Punjab.
(Demo GIF placeholder — imagine a sleek, fast-paced walkthrough of the UI filtering influencers and generating live ML scores)
| 🧠 Contextual AI Matching | 🛡️ Fraud Detection Engine | 📈 Live Audience Telemetry |
|---|---|---|
Uses Sentence-BERT transformer models to semantically map brand requirements to creator profiles. |
Live scoring algorithms detect engagement anomalies and flag suspicious bot-inflated audiences. | Direct ingestion of the YouTube Data API v3 for real-time engagement and subscriber metrics. |
| 💡 Explainable AI (XAI) | 🌍 Dynamic Localization | ⚡ High-Speed Execution |
|---|---|---|
Generates human-readable reasoning ("Why this influencer?") for every algorithmic decision to build trust. |
Zero-latency, client-side currency conversions (USD, INR, EUR, GBP) using reactive JavaScript. | Pre-computed embeddings enable sub-100ms vector similarity searches across the corpus. |
| Campaign Setup (V2) | AI Match Results |
|---|---|
|
 |
| Refined campaign setup with regional state filtering and multi-platform sync capability. | Ranked, explainable AI results showing exact cosine similarity scores and fraud risk labels. |
| Analytics Dashboard | AI Negotiation Script |
|---|---|
 |
 |
| Real-time visualization of creator ROI, engagement metrics, and campaign budget forecasting. | Automated AI drafting of personalized creator outreach emails based on match metrics. |
| Influencer Registration | Influencer Campaign Offer |
|---|---|
 |
 |
| Streamlined onboarding for creators to join the verified InfluenceIQ talent pool. | Personalized dashboard for creators to view and accept brand invitations. |
| Influencer Offer Generation | Secure Authentication |
|---|---|
 |
 |
| Seamless campaign offer structuring and pricing recommendations. | Glassmorphic, enterprise-grade secure authentication entry point. |
| Creator Settings | |
|---|---|
 |
|
| Persistent creator configurations and theme tracking. |
InfluenceIQ is built on a scalable, decoupled architecture pattern, separating the high-speed Flask API gateway from the heavy Machine Learning inference engine.
graph LR
Client([💻 Web Client]) <-->|REST API JSON| Gateway{Flask Gateway}
subgraph Backend Infrastructure
Gateway --> Auth[JWT Auth Service]
Gateway --> ML[🧠 ML Inference Engine]
end
subgraph External Data
ML <-->|Live Sync| YT[(YouTube API)]
ML <-->|Pre-loaded| DB[(Verified Creator DB)]
end
style Client fill:#2b3137,stroke:#fff,color:#fff
style Gateway fill:#0366d6,stroke:#fff,color:#fff
style ML fill:#2ea44f,stroke:#fff,color:#fff
style YT fill:#d93f0b,stroke:#fff,color:#fff
graph TD
A[Brand Input: Budget, Niche, Platform] --> B(NLP Preprocessing)
B --> C{Embedding Engine}
C -->|Fallback| D[TF-IDF Vectorizer 256-dim]
C -->|Primary| E[Sentence-BERT 384-dim]
D --> F((Cosine Similarity Matrix))
E --> F
F --> G[Heuristic Weighting Engine]
G -->|Applies Fraud Penalty & Reach Bonus| H[Final Suitability Score]
H --> I([XAI Reason Generator])
style C fill:#6f42c1,stroke:#fff,color:#fff
style F fill:#e3b341,stroke:#fff,color:#000
style H fill:#2ea44f,stroke:#fff,color:#fff
InfluenceIQ is built for rapid local deployment and seamless cloud scaling.
# Clone the repository
git clone https://github.com/JiveshCodes/InfluenceIQ.git
cd InfluenceIQ
# Initialize virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activatepip install -r requirements.txtCreate a .env file in the root directory:
FLASK_ENV=development
SECRET_KEY=your_secure_random_hash
YOUTUBE_API_KEY=your_google_cloud_youtube_keypython app.pyApplication runs on http://localhost:5000
For enterprise scaling, use the provided gunicorn WSGI configuration.
Start Command:
gunicorn -w 4 -b 0.0.0.0:$PORT app:appNote: Ensure all .env variables are securely injected into your cloud provider's Secret Manager.
- v1.0 — TF-IDF Heuristic Matching & Basic Flask Gateway
- v2.0 — Sentence-BERT Integration, Live YouTube Sync, and XAI
- v2.5 — PostgreSQL Migration & Secure JWT User Authentication
- v3.0 — Vector Database (Pinecone) Implementation for O(1) latency
- v4.0 — LLM-Powered Chatbot for conversational brand interactions
Experience the live version of Influence IQ here:
🔗 https://influenceiq-r4yw.onrender.com/
Note: The application is hosted on Render's free tier, so the initial load may take a few seconds if the server is inactive.
- Modern and responsive UI
- AI-powered functionality
- Interactive user experience
- Real-time processing and dynamic responses
- Optimized frontend design
This project is deployed using:
- Frontend / Backend Hosting: Render
- Version Control: GitHub
Visit the live application to explore all features in action.
- All API keys and cryptographic salts are strictly isolated via
.env. - Incoming JSON payloads are sanitized to prevent prompt-injection and overflow errors.
- Stateless architecture ensures safe horizontal scaling across worker nodes.
We welcome elite engineering contributions. Please fork the repository, adhere to Conventional Commits, and open a detailed Pull Request outlining your architectural changes.
