about / README.md
asansanwal's picture
Update org profile card with full content and links
8ff3556 verified
|
Raw
History Blame Contribute Delete
4.78 kB
---
license: other
tags:
- iab
- adtech
- content-classification
- web-intelligence
- brand-safety
- multilingual
pretty_name: QuickPod AI Web Intelligence Platform
---
<div align="center">
# 🚀 QuickPod AI
**Enterprise AI infrastructure · Web intelligence · Content classification at scale**
<br/>
[![quickpod.io](https://img.shields.io/badge/🌐_quickpod.io-0ea5e9?style=for-the-badge)](https://quickpod.io)
[![quickaihub.io](https://img.shields.io/badge/🤖_quickaihub.io-6366f1?style=for-the-badge)](https://quickaihub.io)
[![aiquick.io](https://img.shields.io/badge/⚡_aiquick.io-10b981?style=for-the-badge)](https://aiquick.io)
[![websitetools.io](https://img.shields.io/badge/🛠️_websitetools.io-f59e0b?style=for-the-badge)](https://websitetools.io)
</div>
---
## Who We Are
QuickPod AI builds production-grade AI pipelines that transform raw web crawl data
into structured intelligence for adtech, brand safety, compliance, and publisher
monetisation.
We operate across the full stack — from distributed crawl processing (Apache Spark
over 100K+ Common Crawl shards) through multilingual neural translation and
ModernBERT fine-tuning, to real-time inference APIs.
---
## Flagship: IAB Content Taxonomy Classifier
Our headline model classifies web content against the **IAB Tech Lab Content Taxonomy 3.0**
across all three hierarchy levels.
| Tier | Categories | Accuracy |
|------|-----------|----------|
| Tier 1 — Broad topic | 27 | **98.1%** |
| Tier 2 — Sub-category | 434 | 79.8% |
| Tier 3 — Specific topic | 217 | 56.1% |
**Training data:** 33.5 million multilingual domain examples · 26 languages
**Architecture:** ModernBERT-base · unified multi-head encoder · hierarchy-consistency loss
**Languages:** ar · bn · de · en · es · fa · fr · hi · id · it · ja · ko · mr · nl · pl · pt · ru · sw · ta · te · th · tr · uk · ur · vi · zh
🔗 **[Try the live demo →](https://huggingface.co/spaces/quickpod/iab-content-classifier)**
---
## Products & Capabilities
| Product | Description |
|---------|-------------|
| **IAB Classifier API** | Real-time URL/text → IAB tier-1/2/3 classification |
| **Domain Authority Index** | CC-MAIN DuckDB authority scores for 200M+ domains |
| **Keyword Intelligence** | Per-domain keyword aggregation from WAT crawl metadata |
| **Multilingual Taxonomy** | IAB classification in 26 languages |
| **Content Safety Models** | Spam · Toxicity · NSFW · Sentiment · Emotion classifiers |
| **Brand Safety Scoring** | GARM-aligned content tier mapping for DSP/SSP |
---
## Use Cases
| Industry | Application |
|----------|-------------|
| **Programmatic Advertising** | Brand-safe contextual targeting, IAB taxonomy compliance |
| **Publisher Monetisation** | Automated inventory categorisation for DSP/SSP integrations |
| **Content Moderation** | Real-time spam, toxicity and NSFW flagging at ingest |
| **Search & Discovery** | Topic classification for crawled and indexed content |
| **Ad Verification** | Pre-bid and post-bid URL classification at scale |
| **Data Enrichment** | Domain-level taxonomy enrichment for audience segments |
---
## Technical Pipeline
Our 8-stage production pipeline:
```
Kaggle (98K domains) + Common Crawl WET/WAT (100K+ shards)
LLM Label Correction (AWS Bedrock — 49.7% label error rate found & fixed)
IAB Seed XL (LLM-synthetic examples for all 678 IAB taxonomy nodes)
Unified Hierarchical Training Data (English, Spark)
Argos Neural Translation → 26 languages, 33.5M rows
ModernBERT Fine-tuning (sharded training, RTX 5090)
Real-time Inference API + HuggingFace Demo Space
```
📋 [Full pipeline documentation →](https://huggingface.co/datasets/asansanwal/wet-iab-ds-multilingual-unified-training-all26-v1/blob/main/PIPELINE.md)
---
## Resources
| | |
|--|--|
| 🔬 **Live Demo** | [quickpod/iab-content-classifier](https://huggingface.co/spaces/quickpod/iab-content-classifier) |
| 📊 **Training Dataset** | [wet-iab-ds-multilingual-unified-training-all26-v1](https://huggingface.co/datasets/asansanwal/wet-iab-ds-multilingual-unified-training-all26-v1) |
| 📋 **Pipeline Docs** | [PIPELINE.md](https://huggingface.co/datasets/asansanwal/wet-iab-ds-multilingual-unified-training-all26-v1/blob/main/PIPELINE.md) |
| 🌐 **Main** | [quickpod.io](https://quickpod.io) |
| 🤖 **AI Hub** | [quickaihub.io](https://quickaihub.io) |
| ⚡ **Platform** | [aiquick.io](https://aiquick.io) |
| 🛠️ **Web Tools** | [websitetools.io](https://websitetools.io) |
---
*For API access, model licensing, or enterprise integration — contact us via [quickpod.io](https://quickpod.io)*