24-week learning path · 168 daily sessions

Build your own AI model
from the ground up.

Learn the foundations of machine learning, deep learning, NLP and Transformers — then build, train, evaluate and deploy your own small language model. Designed for one hour per day.

⏱ 1 hour / day📅 24 weeks🧠 From scratch🌐 Multilingual focus
COMPLETED SESSIONS
0 / 168
Keep the streak going
OVERALL PROGRESS
0%
CURRENT WEEK
Week 1
Based on your progress
PROJECTS
0 / 24
Weekly practical milestones

Your learning overview

Start with the current week. Each session has a topic, task and learning links.

Phase 01 · Weeks 1–5

Foundations

Python, data, linear algebra, calculus and probability.

Phase 02 · Weeks 6–12

Machine Learning & Deep Learning

Models, losses, optimizers, neural networks and PyTorch.

Phase 03 · Weeks 13–19

NLP & Transformers

Tokenizer, language modeling, attention and GPT.

Phase 04 · Weeks 20–24

Training & Production

Dataset pipeline, training, optimization, evaluation and deployment.

Continue learning

Your next incomplete session is automatically selected.

24-week roadmap

Expand a week to view daily learning tasks and resources.

Learning resources

Recommended official and hands-on resources. Open the link and study the relevant section.

Milestones & projects

Every week ends with a tangible artifact. These projects build your portfolio.

Settings & data

Your progress is stored locally in this browser.

Export backup

Download a JSON file with completed sessions and notes.

Import backup

Restore your progress from a previously exported JSON file.

Reset progress

This clears all completed sessions and notes from this browser. Export a backup first.

About this roadmap

One-hour daily schedule: 15 minutes theory, 35 minutes coding, 10 minutes review. It is a learning plan, not a promise that a production-scale LLM can be trained in 24 weeks. The final project is a small research model that demonstrates the full pipeline.