If I were a software intern today
What would I bring to a software internship interview today?
One small project I could explain without opening an AI chat. Software Mansion's internship guide says applicants should describe a project they're proud of and explain which part they built. A personal project counts. So does a team project when you can speak clearly about your contribution.
I work in Kenya, where a good first opportunity can draw a lot of applicants. AI makes it easier to produce a polished demo. I want the project to hold up when somebody starts asking how it works and what happened when something went wrong.
Start with somebody else's boring task
Forget the idea of building a whole company as a portfolio project. Find one person who does the same fiddly task every week. Ask them to show you. Maybe they check a spreadsheet for missing product codes. Maybe they copy numbers between two reports. Watch once, then ask where mistakes usually happen.
Take the spreadsheet example. Claude can write a script to highlight empty cells in seconds. But what does an empty product code mean? A new item waiting for approval? A mistake in the source file? A code that was never assigned? The person doing the work knows. Their answer decides what the tool should flag.
Now make a sample file with invented rows. Build the smallest thing that separates the cases. It could be a spreadsheet formula or a short program. Show the person the output and ask what you got wrong. Fix one thing they point out. There is your project: a real task, a working attempt, and a change made after feedback. Use company data only if you have permission.
Imagine a warehouse app where a scan leads to a document in an inventory system. The code that sends the request may be straightforward. The difficult questions are what staff are scanning, which document should be created, and what to do when a request fails. A small student project will have fewer moving parts, but the habit of asking about the work before automating it is the same.
Make AI earn its place in the project
Use it to explain a CSV parser. Ask for a first test. Let it show you another way to write a function. Then run the code yourself. Feed it a file with a missing column. Change a rule without asking it to make the edit. If you cannot tell why the output changed, stop and work that out before adding features.
There is a reason to be deliberate here. In a small 2026 study, developers learning a new Python library scored lower on a follow-up test when they had AI assistance. The researchers also observed that people who asked questions to understand the work tended to do better than those who handed the task over. That finding will not predict every person's experience. It does make me want to ask a simple question of my own projects: can I explain the part AI wrote?
Software Mansion's engineering internship expects programming basics. A project gives you somewhere to practice those basics: read an error, use Git, trace what happens to an input, and write a test for the case that broke last time. You can learn them while building. You still need to learn them.
What a team can learn from one small project
Software Mansion gives interns real project work early. I like that approach. A beginner gets to work on something somebody needs and see how experienced engineers handle the details.
A new developer could practice against a mock server instead of a live business system. Give them sample data, a clear expected result, and room to make mistakes before their work touches real records.
On a team, people with different levels of experience can review one another's work. An intern should get to show an early version, hear what is wrong with it, and try again. An applicant can ask about that: What would I work on first? Who can answer questions about it? When will someone look at an early version? Those answers tell you more about the opportunity than a long list of frameworks.
Take it into the interview
The spreadsheet checker, the sample file, and a short explanation. Here is the task I watched. Here is the rule I got wrong at first. Here is the feedback that changed it. Here is one input the tool still cannot handle. Here is where AI helped, and here is the part I changed myself.
You can have that conversation without pretending to be an expert or having a company name on your CV. It may not get you the internship. There are more applicants than places, and interviews test other skills too. But it gives someone a real piece of your thinking to examine.
If I were starting over, that is what I would build: one small tool for one person, with enough understanding behind it to answer the next question.