Pathway 1: "I Want to Build AI Products" (Product Manager / Startup Founder)

Target audience: Product managers, startup founders, business leaders building AI-powered products

Goal: Understand what LLMs can and cannot do, how to build products around them, and how to evaluate cost, quality, and risk.

Chapter Guide

• Skip Ch 00: ML and PyTorch Foundations you won't write PyTorch code
• Skip Ch 01: NLP and Text Representation NLP theory not needed for product work
• Skip Ch 02: Tokenization and Subword Models tokenization internals are abstracted away
• Skip Ch 03: Sequence Models and Attention attention math not required at this level
• Skip Ch 04: The Transformer Architecture architecture details handled by providers
• Skip Ch 05: Decoding and Text Generation decoding logic lives behind the API
• Skim Ch 06: Pre-training and Scaling Laws understand cost vs. capability tradeoffs
• Skim Ch 07: The Modern LLM Landscape know which models fit which use cases
• Skim Ch 08: Reasoning Models and Test-Time Compute learn when reasoning models add value
• Skim Ch 09: Inference Optimization understand latency and cost levers
• Skip Ch 18: Interpretability interpretability is research-oriented
• Focus Ch 10: Working with LLM APIs your primary integration point with LLMs
• Focus Ch 11: Prompt Engineering master the prompts that drive your product
• Focus Ch 12: Hybrid ML+LLM Architectures combine classical ML with LLM strengths
• Skip Ch 13: Synthetic Data synthetic data is a training concern
• Skip Ch 14: Fine-Tuning Fundamentals fine-tuning is beyond product scope
• Skip Ch 15: PEFT adapter methods are for ML engineers
• Skip Ch 16: Distillation and Merging distillation is a training specialty
• Skip Ch 17: Alignment alignment research, not product work
• Skim Ch 19: Embeddings and Vector Databases understand how retrieval powers search
• Focus Ch 20: RAG ground your product in real data
• Skim Ch 21: Conversational AI learn dialogue patterns for chat features
• Focus Ch 22: AI Agents agents will power your next features
• Skim Ch 23: Tool Use and Protocols know how tools connect to your product
• Skip Ch 24: Multi-Agent Systems multi-agent complexity rarely needed early
• Skip Ch 25: Specialized Agents specialized agents are an advanced topic
• Skim Ch 26: Agent Safety and Production understand guardrails before shipping
• Skim Ch 27: Multimodal Models know what multimodal unlocks for users
• Focus Ch 28: LLM Applications real-world application patterns and recipes
• Focus Ch 29: Evaluation and Experiment Design measure quality before you ship
• Skim Ch 30: Observability and Monitoring track cost, latency, and errors in production
• Focus Ch 31: Production Engineering deploy reliably and scale with confidence
• Focus Ch 32: Safety, Ethics and Regulation legal and ethical risks you must manage
• Focus Ch 33: Strategy, Product and ROI ROI frameworks and pricing strategy
• Skim Ch 34: Emerging Architectures stay ahead of emerging capabilities
• Skim Ch 35: AI and Society societal context for responsible products
• Focus Ch 36: From Idea to Product Hypothesis AI product framing, model role assignment, feasibility assessment
• Focus Ch 37: Building and Steering AI Products observe-steer loops, prototype iteration, AI coding assistants
• Focus Ch 38: Shipping and Scaling AI Products launch economics, provider portability, post-launch monitoring