Analyzing production code with AI

Advanced AI-Assisted Coding for Developers

Thalia Barrera

AI Engineering Curriculum Manager, DataCamp

Beyond code generation

nanobanana: full: analyzing production code with AI

Course overview

nanobanana: half: developer at desk with multiple monitors

During this course, you will learn to:

✔ Perform AI-based code analysis for production code
✔ Apply AI-assisted testing and security
✔ Use AI for software design and collaboration

The scenario: Wayfarer Labs

nanobanana: half: tourism startup office with world maps

💻 You are a software developer at Wayfarer Labs
🌎 A tourism startup building a mobility platform
🔎 Your task: find an open-source analytics tool

Discovering Atlas

Key questions before adopting:

Is it maintainable?
Is it reliable?
Is it a solid long-term addition?

🤖 AI can help accelerate and systematize the evaluation of code!

Understanding a new repository

nanobanana: full: AI guiding exploration of a new repository

Structured prompting for codebase orientation

🤖 Static analysis prompt:

The following repository contains a Python analytics toolbox called Atlas, intended for use in a production environment.

Assume I have access to a Unix-like system and can install common open-source tools.

I want to perform a static analysis of this codebase without executing it.

Your goal is to:

Recommend appropriate static analysis tools for this project

Explain what each tool measures and why it is relevant

Provide concrete commands to run the analysis

Explain how to interpret the key results

AI-guided dynamic analysis

nanobanana: full: dynamic analysis profiling

Assessing code quality with AI

🤖 Maintainability prompt:

Analyze the codebase for maintainability concerns: identify overloaded modules, functions with blurred responsibilities, and areas where a single change would require edits across multiple files.

Maintainability issues:

Overloaded modules
Blurred responsibilities

Complexity and technical debt

🤖 Complexity prompt:

Identify areas of high logical complexity, focusing on long functions, nested conditionals, and branching.

Complexity indicators:

Long functions
Nested conditionals
High branching

🤖 Technical debt prompt:

Look for signs of technical debt, such as TODOs, duplicated logic, or legacy patterns.

Technical debt signals:

TODOs and FIXMEs
Duplicated logic
Legacy patterns

Evaluation conclusion

Verdict on Atlas:

✅ Well-structured codebase
✅ Reasonable to adopt
👀 Address complexity issues first
👀 Tackle technical debt

AI-assisted refactoring

AI can also helps us:

Refactor focusing on high-impact changes
Propose patches

Important reminder

nanobanana: half: developer reviewing AI suggestions critically

AI is not a replacement for engineering judgment

It can make mistakes
It can miss context
It can hallucinate

✔ Used correctly, it accelerates understanding and helps us make better coding decisions

Let's practice!

Advanced AI-Assisted Coding for Developers