⚡️ 7-Day AI Engineering Sprint

Welcome to the 7-Day AI Engineering Sprint. This repository contains the source code, scripts, and educational materials for building production-grade AI systems from scratch.

Each day focuses on a critical component of the modern AI stack, moving from theory to production engineering.

📅 Curriculum

Day 1: The Foundation

Theme: Deep Dive into AI Engineering

• 📂 Source: src/render_day1_deep_dive.py
• 🎯 Goal: Setting up the environment, understanding the landscape, and preparing for the sprint.

Day 2: Transformers from Scratch

Theme: The Architecture that Changed Everything

• 📂 Source: day2_transformer/
• 🎯 Goal: coding a Transformer model from scratch (no PyTorch/TensorFlow) to understand the math.

Day 3: Self-Attention Mechanism

Theme: The "Brain" of the Transformer

• 📂 Source: day3_attention/
• 🎯 Goal: Visualizing and implementing the Query, Key, Value attention mechanism.

Day 4: Streaming & Latency

Theme: Real-Time AI Systems

• 📂 Source: day4_streaming/
• 🎯 Goal: Building low-latency streaming pipelines with FastAPI and WebSockets.

Day 5: Production RAG

Theme: Retrieval Augmented Generation

• 📂 Source: day5_prod_rag/
• 🎯 Goal: Building a robust RAG pipeline with hybrid retrieval (Vector + BM25) and grounding.

Day 6: RAG Evaluation (The Judge)

Theme: "But would you ship it?"

• 📂 Source: output/rag_eval_github/ (Release Package)
• 🎯 Goal: Building an automated "Judge" system to evaluate RAG accuracy using a Golden Dataset.

Day 7: The Guardrails (Upcoming)

Theme: Security & Safety

• 🎯 Goal: Prompt injection defense, firewalls, and making your AI hack-proof.

🛠️ Usage

This repository is structured as a Content Engine.

• src/: Contains video rendering scripts (Manim/MoviePy) for generating the educational content.
• dayX_*/: Contains the standalone code examples for each day.

To run the RAG Judge from Day 6:

cd output/rag_eval_github
pip install -r requirements.txt # (if available) or pip install openai
python rag_judge.py

📺 Follow the Sprint

Subscribe to follow the daily releases and deep dives.