Yuan's Blog
EN 中

Core Skills and Interview Standards for the Next Generation of GenAI Application Engineers — Andrew Ng

Dear friends,

With the rise of generative AI, a new type of developer has emerged — GenAI Application Engineers — who can build powerful applications at unprecedented speed. These talents are in high demand across industries, though the job descriptions are still evolving. Below is my summary of the key skills such engineers need, and the questions I use during interviews to identify them.

🧠 Two Core Abilities of a GenAI Application Engineer

  1. Ability to build powerful applications quickly using AI building blocks
  2. Ability to develop systems rapidly with the help of AI tools — far faster than traditional development

At the same time, having strong product and design intuition is a huge bonus.

🧱 What Are AI Building Blocks?

Just like LEGO pieces of different shapes can form complex structures, the more AI building blocks you have, the more freely you can combine them into powerful systems.

Common AI Building Blocks Include:

  • Prompt engineering techniques
  • Agentic frameworks
  • Evals (model evaluation tools)
  • Guardrails (safety frameworks)
  • Retrieval-Augmented Generation (RAG)
  • Voice interaction stack
  • Asynchronous programming
  • Data extraction, embeddings, and vector databases (vectorDBs)
  • Model fine-tuning
  • Graph databases + LLMs (graphDB + LLM)
  • Agentic browser / computer control
  • Multi-component reasoning models (MCP, reasoning models)

👉 The more types of building blocks you master, the more complex and intelligent your systems can be.

💡 Although new modules keep emerging, many building blocks from a year or two ago (like vector database frameworks) are still highly useful.

💻 AI-Assisted Programming: Tools Evolve Rapidly

From GitHub Copilot’s debut in 2021 to the emergence of AI IDEs like Cursor and Windsurf, coding assistants have been evolving at lightning speed.

Today, even more advanced AI coding assistants have appeared:

  • OpenAI Codex
  • Anthropic Claude Code (my personal favorite — it can automatically write, test, and debug code across multiple iterations)

🧠 To truly use these tools effectively, you need:

  • A solid foundation in AI and system architecture
  • Clear product goals
  • And not just “vibe coding” (coding aimlessly with AI)

🎯 Combining these tools with strong engineering intuition leads to unprecedented speed and efficiency.

📉 Why Do AI Programming Tools Have a Shorter “Shelf Life” Than Building Blocks?

  • New tools emerge constantly — a Darwinian competition
  • Engineers rarely use dozens of tools at once
  • Each new generation of tools is far more powerful than the last

Therefore, keeping up with the latest AI programming tools is absolutely worth it.

🎨 Bonus: Product Thinking and User Intuition

In some teams, engineers only code according to pixel-perfect design specs. But when every detail depends on a product manager’s instruction, progress slows dramatically.

Now, the shortage of AI product managers makes this inefficiency even worse.

If a GenAI engineer has user empathy and product design intuition, then even with a simple brief (like “create a page where users can view and edit their password”), they can:

  • Make proactive decisions
  • Rapidly build prototypes
  • Drive iterations forward

🧪 Interview Tips: How to Identify a Strong GenAI Engineer?

I usually ask questions in these three areas:

  1. Which AI building blocks are you familiar with? Can you give examples?
  2. How do you use AI programming assistants?
  3. Do you have product intuition or user empathy?

The most important question is:

How do you stay up to date with the latest developments in AI?

Strong candidates typically answer:

  • By subscribing to technical newsletters (like The Batch)
  • Taking short courses or doing hands-on projects
  • Joining discussion communities

Meanwhile, candidates who rely mainly on social media for AI news often struggle to keep up.

🛠️ Keep Building!

Mastering AI building blocks + using AI coding tools effectively + having product intuition — this is the golden combination for the software engineers of the future.

Keep building! — Andrew Ng

📎 Original Article: The Batch — Issue 305