Arc

Arc is an AI foundations portfolio project.

The goal is not to wrap an AI API first. The goal is to build the core ideas behind modern AI in small, visible chapters: vectors, training, tokenization, embeddings, attention, transformers, RAG, and agent loops.

Each chapter should leave behind three things:

• a small implementation
• a visual demo or screenshot-worthy output
• a short note explaining what was built, what worked, and what failed

Structure

apps/
  Final applications built from the chapters.
chapters/
  Grouped AI learning chapters.
docs/
  Roadmaps, notes, and portfolio writeups.

Chapter Map

chapters/
  chapter_01/  # Foundations
    vectors/
    tiny_ml/
    neural_network/
  chapter_02/  # Language Models
    tokenization/
    embeddings/
    attention/
    transformer/
  chapter_03/  # AI App Systems
    rag/
    evals/
    structured_output/
    advanced_rag/
  chapter_04/  # Agentic Systems
    agent_loop/
    agent_patterns/
    mcp_tooling/
    observability/
  chapter_05/  # Model Improvement
    post_training/
    inference_optimization/
    data_centric_ai/
  chapter_06/  # Multimodal AI
    multimodal_embeddings/
    vision_audio/
    diffusion_intro/
  chapter_07/  # Safety and Reliability
    ai_safety_security/

Portfolio Rule

Every chapter must be understandable from its README and one visual result. If a recruiter cannot see what was learned and what was built within 30 seconds, the chapter is not finished.