What to learn to become an AI engineer

6 min read

“Software engineering is dead. Frontend is dead. Backend is dead.”

You must be hearing this every day, either through influencers or AI frontier lab founders.

But those headlines miss the context. The work is changing. We still need people who can understand a problem, design a system, and build software that works.

What’s changing is the technology we use and what we’re expected to build with it.

The IT industry has always been like this. If you have been in the industry for more than 10 years, then you’ve seen this before with the web, mobile, and cloud computing. Some skills became less valuable, others became more important, and new roles emerged.

Now, AI is driving another shift.

So if you’re a developer or someone learning to code, the question is: what should you learn to build useful applications with AI?

A linocut of a person walking a winding path from a desk, past a gem, papers, and a checklist, toward an open door

Understand what an AI engineer actually does

The title can mean different things at different companies.

For an AI engineer, you might build a tool that answers questions about company documents, helps customer support teams, or automates part of a business workflow.

For an AI engineer, you don’t need to begin by training a large model from scratch or mastering advanced mathematics. You do need to understand what models can do, where they fail, and how to build reliable software around them. (though if this is your area of interest, there is another part of AI you can explore and work in)

As an AI engineer, your job goes beyond getting a model to produce an impressive answer. You need to make that capability useful inside a real product.

Build your software engineering foundations

I know, you must be interested in: what should you learn to be an AI engineer?

  1. Start with a language you can use confidently, or treat this as a pivot (a golden opportunity) and learn a new one.

  2. Python is a practical choice for AI development. If you already work with JavaScript or TypeScript, you can also use those skills to build AI applications.

  3. Then get comfortable with Git, the CLI, HTTP, APIs, JSON, and environment variables.

  4. You should be able to call an API, handle an error, protect an API key, and work with a database.

These might sound like ordinary software engineering skills. They are, and they still matter. Hence, many engineers (not influencers) keep saying - "Do not leave coding, do not ignore software engineering fundamentals."

An AI application needs authentication, data storage, testing, and error handling just like other applications.

If you already know these things, you’re building on an existing foundation.

As an AI engineer, your job is not just giving a "prompt" to an AI agent. Your job is to guide AI on what to build, how to build it, and how to design and maintain workflows.

Understand how to work with AI models

There is a HUGE difference between "vibe coders" and "AI engineers". Apart from "software engineering fundamentals", it is important for AI engineers to learn about AI and how to work with AI models.

  1. Start with tokens. Tokens are the units a model processes. They affect how much information you can send and how much a request can cost.

  2. Context: The context window is the amount of information a model can work with in a request. You need to decide what belongs in it.

  3. Prompting: how to give clear instructions, provide relevant examples, and specify the output you need.

  4. Structured outputs and tool calling: These help connect model responses to application code and external systems.

  5. Retrieval-augmented generation, or RAG: This means retrieving relevant information such as passages from your documents and providing it to the model when it answers.

  6. Memory: A model doesn’t automatically remember everything about a user. Your application needs to decide what to store, retrieve, update, and forget.

Once those foundations make sense, explore MCP, which provides a standard way for AI applications to connect to tools and data sources.

You don’t need to learn all of these concepts in one weekend. Learn each one when your project gives you a reason to use it.

Learn how to evaluate, deploy, and maintain the application

But how does a working demo become a useful product? Start with evaluation. How will you know whether the application is doing a good job?

Create a small set of realistic questions and examples. Include difficult cases and questions the application should decline to answer. Check accuracy, relevance, and whether its answers are supported by the information you supplied. Then learn deployment and observability.

  1. Can you see which requests failed?

  2. How long did they take?

  3. What did they cost?

  4. What information did the model receive?

You also need to think about privacy, access controls, and what happens when a document or user input contains instructions that try to redirect the application.

And whenever you change a prompt, model, or retrieval strategy, run your evaluations again.

A change that improves one example might make several others worse.

So, what should you build first?

You can start by building a small application that answers questions about a set of documents. Start by calling a model through an API. Then add document retrieval, show the sources behind its answers, and make it say when the documents don’t contain enough information.

Create a small evaluation set. Track errors, latency, and cost. Finally, deploy it so someone else can use it.

This one project gives you practice with APIs, prompting, context, RAG, evaluation, and deployment.

You can add tools or agents later, when there’s a clear reason.

Becoming an AI engineer doesn’t mean abandoning software engineering. It means extending your skills to work with systems whose outputs aren’t always predictable.

Start with a real problem. Build a small solution. Measure how well it works. Then improve it.

That’s the roadmap. That's how you can call yourself an "AI Engineer".

A few projects I built (take inspiration)

https://github.com/Neha/jev-vs-openai-decisions

https://github.com/Neha/apartment-4a

https://github.com/Neha/go-learn-app

https://github.com/Neha/Solaris

https://github.com/Neha/ai-trends-agent

https://github.com/Neha/check-fix-accessibility

https://github.com/Neha/rn-developer-skills

https://github.com/Neha/loop-engineering-demo

Happy Learning!!

← Back to home