The question of “What makes someone a senior level software engineer?” is one that I have been pondering for a while. Whatever the answer was a year ago, it is different now. These are my thoughts today, based on my experience using AI and from what I can gleam from others.

AI Literacy

This should be obvious. The ability to use AI effectively to do things like evaluate requirements, scope work, write code and create tests cases is a requirement.

Application Architecture

AI is very good at coding. It is not yet as good at application architecture. A main criterion for a well-structured application is that it is easy to maintain. A well-structured application is not just easier for a human to understand and modify; it is easier for AI also. There is a lot of bad code and poorly constructed applications in the world. I would say more bad than good because all of us have written bad code at some point but not everyone progresses. AI does not appear to know the difference between good and bad.

Broad Experience

Now more than ever broad experience is more important than deep experience. The different technology ecosystems tend to become echo chambers where each has a single blueprint for how applications ought to be structured based on the type off applications that were historically built using those tools, and AI is likely to follow that. But there is no architecture, structure, tool that is best for every situation. Engineers with exposure to more ecosystems are much more aware of the alternatives.

Problem Solving

Ultimately the purpose of of software engineering is problem solving. It may be a bug in existing code. I may be finding a solution to a business process problem that would be both effective, usable and likely to be adopted. It may be how to provide a strategic advantage for a business. Sometimes the problems are complicated, and the solutions are not obvious. AI can often suggest solutions, but even those cases, we need to be able to evaluate whether it is the best solution.

Strategic Thinking

It is still an important part of the role to deliver high-impact outcomes by evaluating cost vs benefit in everything that we do. We still need to evaluate risk. We still need to be proactive.

Data Modeling

AI is probably going to be creating our data models at least in some cases. But we still need to be able to review the models to verify that they are correct. AI can easily misunderstand requirements, and these misunderstandings will often show up in the data model. Applications that are built on top of a flawed data model are difficult to fix particularly if they are in production. Do organizations with multiple locations? Will we allow multiple admin users per account? How are privileges assigned? Do we need to track batches or lots for our inventory in case of a recall? Do we need recurring payments? Do we allow multiple email addresses? These all impact the data model.

Security Best Practices

Particularly if in high security fields like health care, financial or national security, I don’t think anyone has confidence that AI will consistently implement adequately secure solutions. The consequences are just too big.

Those are my thoughts today. In six months I may have a different answer. I am interested in hearing yours.