Projects Index

Hands-on projects to solidify learning. Code lives in ../projects/, notes here.

Beginner (Phase 1-2)

• EDA on a dataset — load, clean, explore, visualize a real dataset
• Linear regression from scratch — implement gradient descent in NumPy
• Iris classifier — scikit-learn workflow: split, scale, train, evaluate
• Titanic survival prediction — feature engineering, multiple models, comparison

Intermediate (Phase 2-3)

• MNIST digit classifier — first neural net in PyTorch
• Sentiment analysis — TF-IDF + logistic regression, then fine-tuned BERT
• Image classifier with transfer learning — fine-tune ResNet on custom dataset
• Customer segmentation — K-means clustering on e-commerce data

Advanced (Phase 3-4)

• Build a RAG system — embeddings + vector search + LLM
• Train a small language model — understand tokenization and training loops
• Object detection — YOLO on a custom dataset
• RL agent — solve CartPole, then a harder environment

Project workflow

1. Define the goal clearly
2. Get data (see ../datasets/)
3. Explore and clean (EDA)
4. Build a baseline (simplest model that could work)
5. Iterate: better features, better models, better tuning
6. Document what you learned in a note here

Links

• AI-ML Home
• MLOps Roadmap