NERV is an applied machine learning project that brings trained TensorFlow models into a practical, deployable workflow.
This repository focuses on using, integrating, and evaluating models that were originally trained and validated in my dedicated TensorFlow learning repository.
As a self-taught programmer, I built this project to challenge myself and bridge the gap between AI theory and web deployment. What started as a month-long deep dive into the CS50/TensorFlow stack is now a functional, deployed application on Railway.
All core models used in this repository were trained in the following TensorFlow-focused repository:
🔗 Training Repository (TensorFlow):
https://github.com/aypy01/tensorflow
That repository documents:
- Data preprocessing
- Model architectures
- Training strategy
- Evaluation metrics
- Saved
.kerascheckpoints
NERV consumes those trained models and applies them in a structured application context.
| Model File | Task | Dataset | Performance |
|---|---|---|---|
titanic.keras |
Binary Classification | Titanic Survival | ~81% Accuracy |
iris_species.keras |
Multiclass Classification | Iris Dataset | ~70% Accuracy |
cifar10.keras (Oculus) |
Image Classification | CIFAR-10 | ~72% Accuracy |
sentiments.keras (Yapper) |
Text Classification | IMDb Reviews | 85.80% Accuracy |
Model architecture, preprocessing, and training details are documented in the TensorFlow repo linked above.
- Titanic Survival Prediction
- Iris Species Classification
- Oculus (Computer Vision)
- Yapper (NLP)
These models demonstrate classical tabular ML workflows:
- Feature engineering
- Normalization
- Dense neural networks
- Proper train/test separation
- Uses the CIFAR-10 CNN model
- Focuses on image-based inference and integration
- Serves as the visual intelligence component of the system
- Powered by
sentiments.keras - IMDb review classification
- Text vectorization + embedding-based neural network
- Designed for real-world text inference scenarios
- TensorFlow / Keras
- Python 3
- Django (application & integration layer)
- HTML/CSS
- JavaScript
- David J. Malan CS50 instructor
- Brian Yu CS50 Web
- TensorFlow Model training & inference
- Django Backend & web integration
- CS50 Computer science foundations
- scikit-learn Classical ML utilities
- Google Colab Experimentation & prototyping
- ChatGPT Debugging, documentation
- GitHub Copilot Code assistance
NERV is not about experimenting blindly.
It is about applying trained intelligence correctly.
Models are treated as:
- Versioned artifacts
- Measurable components
- Replaceable, improvable systems
Training is iterative.
Deployment is deliberate.
Created and maintained by
