Module 7: AI Engineering - The New Role

1. Module Introduction: A Role That Didn't Exist 3 Years Ago

Next →

Description

Welcome to Module 7: AI Engineering - The New Role. This is a special module because it's about the very role you're preparing to take on.

A surprising reality: The term "AI Engineer" practically didn't exist before 2020. It emerged with the explosion of LLMs and AI APIs (GPT-3, Claude).

In this module you'll understand:

  • What is an AI Engineer? (and what they are NOT).
  • How do they differ from an ML Engineer and a Data Scientist?
  • What skills do you need? (and which ones you do NOT).
  • What do they do day to day?
  • Is it for you? (an honest self-assessment).

Goal: Total clarity about the role, so you make informed decisions about your career.

Why This Role Didn't Exist Before

2010-2019: Only ML Engineers and Data Scientists

The landscape:

  • If you wanted to use AI, you had to train models from scratch (or fine-tune pre-trained models).
  • Only large companies (Google, Facebook, Amazon) could do it (cost, talent, infrastructure).
  • Roles: ML Engineer (they trained models), Data Scientist (they analyzed data).

2020: GPT-3 Changes the Game

OpenAI launches the GPT-3 API (June 2020):

  • For the first time, any developer can use a state-of-the-art LLM without training it.
  • You pay per use (per token), without GPUs or ML expertise.

Result:

  • An explosion of startups that integrate AI (they don't train from scratch).
  • A need emerges for a role that integrates existing models (not one that trains them).

2022-2024: The Explosion of LLMs and APIs

Timeline:

  • 2022 (Nov): ChatGPT → 100M users in 2 months.
  • 2023: Claude 3, Llama 2, Mistral, GPT-4 → a proliferation of options.
  • 2024: OpenRouter, Together.ai, Replicate → aggregators make access easier.

Result:

  • Companies need to integrate AI into products (chatbots, RAG, agents, classification).
  • They need developers who know how to integrate APIs, design systems, optimize prompts → the "AI Engineer" is born.

What an AI Engineer Is (An Initial Definition)

AI Engineer: A developer who integrates AI models (mainly LLMs) into applications.

Characteristics:

  • They don't train models from scratch: They use APIs (OpenAI, Anthropic) or open-source models (Llama 3).
  • They do integrate models: They connect LLMs to applications (chatbots, RAG, agents).
  • They do design systems: The architecture (embeddings, vectors, prompts, orchestration).

Why It's Different from an ML Engineer

ML Engineer (2010-today):

  • Trains custom models (PyTorch, TensorFlow).
  • Advanced math (linear algebra, calculus, optimization).
  • Training infrastructure (GPUs, distributed training, MLOps).

AI Engineer (2020-today):

  • Integrates existing models (APIs, open-source).
  • Minimal math (they understand the concepts, but don't implement them).
  • Inference infrastructure (APIs, vectors, prompts).

Analogy:

  • ML Engineer: Builds the car's engine (from scratch).
  • AI Engineer: Integrates the engine (already built) into the car (the application).

Why This Role Is Valuable

1. Explosive Demand

Statistics (2023-2024):

  • A 300% increase in job postings with "AI Engineer" (LinkedIn).
  • Companies are looking for AI Engineers more than ML Engineers (a 3:1 ratio in startups).

Reason: Integrating AI is more common than training from scratch (OpenAI, Anthropic do the training).

2. A Lower Barrier to Entry

ML Engineer:

  • Requires a PhD or Master's in CS/ML (usually).
  • 2-5 years of experience in ML.
  • Advanced math.

AI Engineer:

  • Requires experience in software development (Python, APIs).
  • Understanding LLMs conceptually (NOT mathematically).
  • 0-2 years of experience (many come in from web development).

3. High Impact

An AI Engineer can:

  • Launch a chatbot in 1 week (vs 6 months with an ML Engineer training a custom model).
  • Integrate RAG in 2 weeks (vs 3 months building a custom system).
  • Add AI features without an ML team (the APIs do the heavy lifting).

What an AI Engineer Is NOT

Clarifications:

  1. They're NOT an ML Engineer: They don't train models from scratch (they use APIs/open-source).
  2. They're NOT a Data Scientist: They don't analyze data or create reports (they build products).
  3. They're NOT a Prompt Engineer: Prompt engineering is a skill, not a complete role.
  4. They're NOT an "AI Guru": You don't need to understand Transformer math (only conceptually).

Why Understanding the Role Matters

1. Correct expectations

Without clarity:

  • You apply to an "AI Engineer" role that is really looking for an ML Engineer (math, training).
  • You study unnecessary advanced math (when you should be studying RAG, agents).

With clarity:

  • You know what to study (LangChain, APIs, vectors, prompts).
  • You know which job postings are appropriate.

2. Career path

An AI Engineer can evolve into:

  • A Senior AI Engineer: Complex architectures (multi-agent, advanced RAG).
  • An AI Architect: Enterprise AI system design.
  • An ML Engineer: If you decide to study training from scratch.

Summary

AI Engineer is a new role (2020-today):

  • It emerged with the explosion of LLMs and APIs (GPT-3, ChatGPT).
  • It integrates existing models (it doesn't train from scratch).
  • It designs systems (RAG, agents, prompts, vectors).

Key differences:

  • ML Engineer: Trains models, advanced math.
  • AI Engineer: Integrates models, minimal math.

Why it matters:

  • Explosive demand (300% growth).
  • A lower barrier to entry (vs ML Engineer).
  • High impact (a chatbot in days, not months).

Next step: Lesson 02: AI Engineer vs ML Engineer — The fundamental differences in detail.

← Previous7. Exercise: Choosing the Appropriate Provider and StackNext →2. AI Engineer vs ML Engineer: Fundamental Differences