I've just started a 20-day summer internship in an R&D department that mainly works on embedded systems, although my section focuses on software rather than electronics. My mentor explained that AI is reducing the time needed for many programming tasks and suggested that I choose a project or combination of topics that would take roughly 15 days and make my internship worthwhile. Web design did not seem to interest him.
I'm entering my third year of study and currently know Java, Bash scripting, C, a little Python, and have recently started learning C# and the .NET framework. I also considered a cybersecurity-related project, but I'm not sure how positively that idea was received.
What kind of project would be challenging and useful enough to take around two weeks, even with AI-assisted development? Ideally, it should relate to the company's work or provide something genuinely useful to the team.
5 Answers
Another practical idea is an analytics or reporting prototype. Find a recurring internal data source, such as test results, build logs, bug reports, or weekly measurements, and create a tool that cleans it up, identifies trends, and produces a useful report or dashboard. That gives you experience with scripting, data processing, and presenting results while producing something the team may actually use.
Rather than inventing a project in isolation, talk with several developers in the department and ask what they’re currently working on. You may find a small unfinished tool, test system, reporting task, or maintenance problem where an intern could make a useful contribution. Once you find something concrete, discuss the scope with your mentor and turn it into a manageable project. Your mentor should ideally have provided a project, so getting context from the team is probably the best way forward.
If the company is interested in AI, consider a small internal proof of concept rather than a full chatbot. For example, you could build a tool that searches approved technical documents, summarizes test logs, or answers questions about internal procedures. The important part would be evaluating accuracy, handling incorrect answers, protecting confidential data, and documenting where the tool should not be trusted.
A C# or .NET backend could be a reasonable option, especially if you include authentication, authorization, input validation, logging, tests, and documentation. Even a small internal service becomes a substantial learning project when you have to make it reliable and secure. Just don’t choose a generic backend without first checking whether it solves a real problem for the company.
AI can speed up implementation, but it doesn’t remove the need to define the requirements, review the generated code, test edge cases, and understand the security implications. Those parts can easily take most of the project time.
Since the department works with embedded software, you could propose a software-only tool around the existing development process: a test-result analyzer, log parser, regression-test runner, configuration validator, or security scanner for internal code. These projects don’t require designing electrical circuits, but they connect directly to the team’s work and can be scoped to a two-week prototype.
That’s how internships are usually structured in my experience: the team has a prepared project and guides the intern through the design and implementation. Someone who is new to the company won’t necessarily know what the team needs yet.