AI-Assisted Software Development — A Course for the Transition Moment

The Context

The ABQ Dialogues #2 session on Romania's tech industry at the crossroads raised a question that remains live in every engineering organisation: what happens to software development roles as AI tools mature? The debate in the room was largely among senior engineers and business leaders. But the more consequential question may be happening one layer down — in the university masters programmes that are preparing the next cohort of engineers to enter the market.

Andrei Oros, a senior engineer at UiPath, contributed a course on AI-assisted software development to approximately 60 students in a master's programme at West University Timișoara. The course ran weekly, with each participant iterating on an individual project throughout the term. The best projects — more than ten of them complete, advanced, and production-quality — were recognised at the end of the programme.

What the Course Tested

The course design was built around a deliberate challenge: could students, equipped with AI tools and a structured weekly iteration process, produce software products that would previously have taken significantly longer or required a larger team? The answer, in the top cohort, was yes. Students produced SaaS and mobile applications that were architecturally sound, feature-complete, and oriented toward real market problems. This was not despite AI assistance but because of how they used it — as a fast prototyping layer on top of their own engineering judgement.

The skill that separated the top performers from the rest was not facility with the AI tools themselves. It was the ability to evaluate what the AI produced, identify what was wrong or incomplete, and make the architectural decisions that the AI could not make. In other words, strong engineering fundamentals remained the prerequisite. The AI compressed the time required to implement decisions — it did not substitute for the capacity to make them.

The Transition Question

A recurring question in the course — and in the broader ABQ Dialogues discussion — is what happens to junior engineers in an AI-augmented market. The immediate concern is real: many of the tasks that previously served as on-ramps for junior developers (boilerplate code, straightforward bug fixes, documentation, basic CRUD endpoints) are increasingly handled by AI tools. This compresses the volume of entry-level work available.

But the transition impact is likely to be temporary rather than permanent. As AI becomes standard organisational infrastructure, the productivity ceiling rises and headcount requirements stabilise — but the profile of the candidate changes. Organisations will want engineers who are proficient with AI tooling, capable of evaluating and directing AI output, comfortable working across disciplines, and able to communicate directly with customers and business stakeholders. The on-ramp to software engineering is not disappearing — it is moving upward.

What This Means for Timișoara's Education System

The ABQ view, supported by the course experience and by the broader survey findings in our West Romania's AI Moment research, is that universities must integrate AI tools into curricula not as a subject of study but as an instrument of practice. Students should work with AI tools while building real projects, under conditions that require them to evaluate, correct, and direct the AI's output. The course at West University is one early example of what this looks like in practice.

The implication for Timișoara's engineering ecosystem is that the quality of the next cohort of engineers will depend substantially on whether universities can redesign project-based learning around AI-augmented development — not as an elective or a specialisation, but as the default mode of practice. The organisations that will win the talent competition over the next decade are those that treat AI proficiency as a baseline, not a differentiator.