This repository contains all completed projects from the ShadowFox Artificial Intelligence & Machine Learning Internship. Each task demonstrates real-world machine learning workflow: dataset handling, preprocessing, model development, evaluation, and deployment-style execution.
ShadowFox/
├── Task1 - CIFAR-10 Image Classification (Deep Learning)
├── Task2 - Car Selling Price Prediction (Machine Learning)
└── Task3 - Gemini Language Model Analysis (NLP & LLM Research)
A deep learning project that trains a custom CNN from scratch to classify images across 10 CIFAR-10 classes such as airplane, automobile, cat, dog, and ship.
Key Highlights
- Custom CNN architecture written in PyTorch
- GPU-accelerated training
- Automatic dataset handling and preprocessing
- Inference script to classify single or batch images
- Model checkpoint saving + reproducible CLI interface
📈 Best Validation Accuracy: 98%
📁 Folder: Task1/
A machine learning project that predicts a car’s selling price using structured automotive data and engineered features such as mileage, fuel type, engine size, and years of service.
Core Features
- Data preprocessing, feature encoding, and transformation
- Multiple regression models trained and compared:
| Model | Included |
|---|---|
| Linear Regression | ✔ |
| Ridge & Lasso | ✔ |
| Decision Tree | ✔ |
| Random Forest | ✔ |
| Gradient Boosting | ✔ |
📈 Best Model: Gradient Boosting
R² Score: 0.9575
📁 Folder: Task2/
A research-focused project evaluating Google’s Gemini language model across multiple NLP dimensions including reasoning, domain adaptability, consistency, and prompt sensitivity.
Analysis Includes
- Multi-turn conversation evaluation
- Creativity and storytelling benchmarking
- Domain-wise LLM performance comparison
- Readability metrics, token analytics, and frequency mapping
- Statistical patterns (correlation, variability, consistency tests)
📊 Sample Results
- Total responses analyzed: 13
- Average response length: 27.3 words
- Context understanding: 100% success rate
- Consistency score: High (CV < 0.1)
📁 Folder: Task3/
| Category | Tools |
|---|---|
| Programming | Python |
| ML/DL Frameworks | PyTorch, Scikit-Learn |
| Data Handling | Pandas, NumPy |
| Visualizations | Matplotlib, Seaborn |
| NLP/LLM | Google Generative AI (Gemini API), textstat |
| Execution Platforms | Jupyter Notebook, Python Scripts |
Throughout the internship, I gained experience in:
- ML workflows from raw data → model deployment style inference
- Training deep learning models from scratch
- Feature engineering and choosing suitable ML models
- Evaluating models using industry-standard metrics
- Working with LLM inference, analysis frameworks, and API-driven NLP systems
- Structuring code, documenting results, and version-controlling projects with Git/GitHub
Each task directory includes:
✔ Source code ✔ Requirements file ✔ Project-specific README ✔ Results (plots, metrics, or checkpoints)
Completed as part of the ShadowFox AI/ML Internship Program. All work is original and executed independently.