Transform raw headlines into actionable insights with a production-ready news classifier that’s fast, accurate, and scalable.

A one-stop solution for automated news filtering in aggregation apps:

• Follows the full ML lifecycle—from data exploration to model deployment
• Targets multi-topic classification using only article titles
• Deployed via TensorFlow Serving for real-time predictions

• Robust Preprocessing: Tokenization, embedding (10k vocab, 24-dim), padded sequences (max length = 20)

• Balanced Experiments: Baseline model (78% accuracy) vs. class-imbalance strategies

• Modular Architecture:

  • lab_utils.py for directory management & data loading
  • e1/ and e2/ folders for isolated experiment runs

• Production-Ready Serving:

  • Docker-friendly TF-Serving commands
  • Built-in support for A/B testing and versioning

1. Clone this repo:

   git clone https://github.com/sohailshk/News_Filtering_InProduction && cd news-classifier

2. Create a virtual environment & install dependencies:

   python3 -m venv venv && source venv/bin/activate
   pip install -r requirements.txt

3. Launch Jupyter & explore:

   jupyter notebook News_Filtering__Orignal.ipynb

1. Train the model: follow 🔢 Preprocessing → 📊 Training → 📉 Evaluation in the notebook

2. Serve your best model:

   tensorflow_model_server \
     --rest_api_port=8501 \
     --model_name=news_classifier \
     --model_base_path=./e2/model/

3. Predict via curl or client call to /v1/models/news_classifier:predict

Why choose this solution?

1. Speed: Lightweight architecture for sub-second inference
2. Accuracy: Proven 78%+ baseline, easily extendable
3. Scalability: Containerized TF-Serving for seamless integration

For partnership, customization, or support, contact [email protected]. Let’s revolutionize how the world consumes news—today!