Kelcode AI Labs

A hands-on AI engineering lab series exploring the full AI lifecycle — from data and fine-tuning to model optimisation, debugging, and deployment. Designed for platform builders and curious engineers, this project blends real-world tools with practical examples across multiple tracks.

Articles supporting this repo can be found at kelcode.co.uk

🧠 What's Inside

This repository supports an end-to-end AI curriculum, structured into four modules:

Module	Focus	Status
1	Foundations of AI Modelling	🔜 Coming Soon
2	Data Science & Modelling	✅ In progress
3	AI Engineering & Deployment	🔜 Coming Soon
4	Advanced AI Platform Arch.	🔜 Coming Soon

For the full details, check out the [CURRICULUM.md] doc

📦 Module 1: Foundations of AI Modelling

• AI lifecycle walkthrough
• Key terminology and system thinking
• Accuracy vs speed vs cost trade-offs

📊 Module 2: Data Science Track (Modelling Focus) [in-progress]

• Data cleaning and exploration
• ML & deep learning models
• LoRA fine-tuning and quantisation (FP16, INT8)
• Evaluation and debugging (ONNX, TensorRT)

⚙️ Module 3: AI Engineering Track (Platform Focus)

• Model packaging (ONNX, TorchScript)
• FastAPI + Triton inference APIs
• Kubernetes deployment via Helm
• Versioning and monitoring

🏗️ Module 4: AI Platform Architecture

• RAG systems and vector DBs
• Multi-tenant architectures
• Security, cost engineering, and MLOps patterns

Each module with it's own set of labs and readmes supporting those particular labs

📁 Directory Structure

kelcode-ai-labs/
├── foundations-ai-modelling/     # Foundational concepts and lifecycle labs
├── data-science-track/           # Modelling-focused labs (LoRA, quant, ONNX, etc.)
│   ├── lab01
│   ├── lab02
│   ├── lab03
│   └── ...
├── ai-engineering-track/         # Engineering labs (APIs, inference, k8s)
├── ai-platform-architectures/    # Architecture and ops
├── CURRICULUM.md                 # Curriculum details, likely to change over time
└── README.md

📚 Articles in the Series

1. Embarking on an AI Engineering Journey: From Data to Deployment
2. Machine Learning: Building Our First Classifier with Scikit-Learn
3. Deep Learning: Training Our First Model
4. Fine-Tuning a Model with LoRA
5. Quantisation in Deep Learning: A Practical Lab Guide
6. Debugging and Repairing Tensorrt Inference
7. Serving Quantised Models with FastAPI
8. Building a Homelab RAG
9. Evaluating a RAG system - Coming Soon!

Contributing

We welcome your feedback and improvements! Please read our Contributing Guide for details on how to report issues, submit pull requests, and follow our commit message conventions.