Redesigning engineering education around AI

The three strategies an institution can adopt are a dedicated AI programme, an interdisciplinary layer on any discipline, and the infusion of AI into existing courses, write Prof B Ravindran and Prof Karthik Raman

TNN| Posted August 20, 2026 01:07 PM

AI should not be treated as a set of electives bolted onto an existing degree, but as a foundational capability that reshapes the engineering curriculum itself the three strategies an institution can adopt toward that end: a dedicated AI programme, an interdisciplinary layer on any discipline, and the infusion of AI into existing courses. Of those three, we take up the dedicated degree here, filling in the details of the AI curriculum and the thinking behind redesigning it from first principles.

BTech in Artificial Intelligence and Data Analytics (AIDA), offered by Department of Data Science and AI at the Wadhwani School of Data Science and AI, was launched two years ago. The curriculum was built from the ground up, rather than as a rebadging of existing computer science courses. Its goals span mathematical foundations, learning algorithms, programming for AI systems, data curation, deployment systems, modelling and simulation, and responsible AI.

The curriculum begins with strong mathematical and scientific foundations including applied linear algebra, probability and statistics for engineers, optimisation, calculus, and courses that connect computing with scientific domains such as Computational Physics, Computational Chemistry, and Computational Biology. It then progresses into a substantial AI core consisting of AI, machine learning I and II, deep learning, online and reinforcement learning, algorithms for data science, MLOps, and responsible AI, accompanied throughout by dedicated laboratory courses, ML Lab, AI Lab, DL Lab, and MLOps Lab, that run alongside the theory.

MLOps should be a core component of the curriculum. It is the discipline of operationalising machine learning through pipelines, versioning, reproducibility, deployment, monitoring, and lifecycle management. Integrating MLOps and its associated labs into the core curriculum can better prepare graduates for the practical demands of taking AI systems from experimentation to production.

AIDA does not simply append AI onto a conventional computer science structure; the CS core itself was reconsidered. Data structures and algorithms remain central and receive dedicated emphasis through two separate courses. Systems coverage is re-weighted toward what a modern data-intensive engineer needs, relatively less time on legacy theoretical material, more on computational thinking, scalable systems, and algorithmic problem-solving. This is a pragmatic shift, not an ideological one.

And because weak data systems sink most real-world AI projects, the AIDA curriculum gives data its due. Courses in data curation and visualisation sit alongside broader data management, spanning relational databases and contemporary paradigms like NoSQL, columnar, distributed storage, and vector databases. In the era of large language models and retrieval systems, data engineering is often as important as the model itself.

Responsible AI is therefore embedded as a core course in the seventh semester, alongside technical subjects. The course covers fairness, accountability, and the legal and ethical dimensions of automated decision-making, reflecting the growing need for engineers to understand not only how AI systems are built, but also their broader implications. This approach is aligned with the wider work on responsible AI at IIT Madras through the Centre for Responsible AI (CeRAI).

The curriculum’s underlying principles are perhaps its most transferable element, offering a useful framework for institutions seeking to build their own dedicated AI programmes.

. Strong foundations before hype. Mathematics, statistics, optimisation, and systems thinking have to precede tool fluency, not follow it.

. Multi-semester depth. Where AI is the major, the foundations, the core, and the labs should span several semesters, rather than being compressed into a single introductory sequence.

. Industry-grade execution skills where deployment, MLOps, hands-on labs, and real projects deserve the same curricular weight as theory, because operating a model reliably is where most graduates fall short.

. Modern data engineering and management. Relational databases, NoSQL and columnar stores, distributed systems, and vector databases are no longer peripheral; in the era of retrieval-augmented systems, the data layer is often as consequential as the model.

. Responsible AI integrated, not appended. Fairness, accountability, and the legal context of automated decision-making belong inside the core, examined curriculum, not as a single late-stage ethics elective.

. Curricular redesign, not cosmetic rebranding. Renaming a course does nothing for the student in the room; the harder work is reconsidering what it actually teaches, and why.

Everything outlined here is within the reach of an institution willing to redesign its AI programme from first principles. The larger question, however, is whether such an approach can extend beyond the limited number of campuses with the autonomy and resources to undertake such a redesign. That is ultimately a question of policy, not curriculum, and one that merits separate consideration. 

(The authors are head, Wadhwani School of Data Science and AI; head, Centre for Responsible AI (CeRAI) and professor, Department of Data Science and AI, and a member of CBSE Expert Committee for the Computational Thinking and AI Curriculum for classes III–VIII respectively)