AI can amplify engineers, but not replace fundamentals, says Gopi

Gopichand Katragadda argues that artificial intelligence is changing the scale and pace of engineering, but not the foundations on which good engineering depends. In an interview reported by The Times of India, he says engineers still need strong technical basics, an understanding of how components work together, and the discipline to build, test and document useful products.

Katragadda has held senior technology roles at GE and Tata Sons. He was managing director of GE’s India Technology Centre in Bengaluru, later became Tata Sons’ first group chief technology officer and innovation head, and now leads Myelin Foundry, an artificial-intelligence company working across media and entertainment, industrial IoT and automotive applications.

His view is that engineering can be understood through three connected areas: theory and practical knowledge; component-level expertise combined with systems thinking; and execution. The last of these includes developing products efficiently, judging whether they will deliver real commercial value, and recording enough information for later teams to understand and improve the work.

Where Indian engineering needs to improve

Katragadda says India has built considerable strength in mathematics, theory and the design of individual components. However, he believes the country needs to do more to develop hands-on engineering, whole-system understanding, research and documentation.

A person may know how to design a turbine blade, he says, without necessarily being equipped to determine how an entire wind farm should be configured. That broader view requires engineers to understand the interaction between separate components, operating conditions and the final system’s purpose.

He also questions the habit of treating low cost as the main measure of successful execution. Developing something cheaply is not enough if the finished product is not commercially viable or does not provide meaningful value. In his view, engineering teams should invest what is necessary to produce work that performs well in the real world.

Documentation is another weak point. Engineers may retain large amounts of knowledge personally, but failing to write it down can make collaboration harder and leave future teams without the reasoning behind earlier decisions. Katragadda sees AI as a possible way to make documentation more consistent, rather than as a replacement for engineers’ own understanding.

Why experience matters in the AI era

Katragadda cautions young engineers against making AI their first source of expertise. Early in a career, he says, people need to learn the underlying domain without relying on shortcuts. Experience with the data, hardware and operating environment is what allows an engineer to ask better questions and recognise whether an answer is sound.

AI can support that learning by explaining difficult theory, helping locate relevant information and suggesting routes toward practical experience. It can also assist with coding, analysis and written records. But the engineer remains responsible for identifying the problem, understanding the important points of failure and deciding what needs to improve over earlier systems.

That responsibility becomes especially important when a system combines hardware and software. Engineers may not know every detail when they begin, but they should be able to study technical documents, investigate faults and test whether an algorithm works on the device for which it was designed. Katragadda’s position is that engineers should not simply accept an AI-generated result; they need to check it and take responsibility for the outcome.

An example from aircraft engines illustrates the value of domain knowledge. Turbine blades contain different types of holes. Some support cooling, while others reduce weight and help improve the thrust-to-weight ratio without undermining structural strength. Understanding why those features exist requires more than generating a plausible explanation: it requires knowledge of the engineering problem and the constraints that shaped the design.

AI applications already in use

Myelin Foundry’s products demonstrate the type of engineering Katragadda believes AI can enable. One of the company’s systems cuts video bandwidth requirements by about 40%. Alternatively, it can enhance low-resolution video for use on modern displays. The bandwidth reduction represents a comparable fall in content-delivery costs, and the system is being used by several international over-the-top streaming companies.

The company has also developed a vehicle rear-view camera system designed to make a night-time image appear much closer to daylight viewing. The aim is to help drivers see more clearly while reversing. Other applications described by Katragadda include detecting anomalies, anticipating maintenance needs and generating documents.

These examples underline his wider point: useful AI engineering is not merely a matter of requesting a generic answer from a model. It involves understanding the physical equipment, software, data and limitations of the environment where the technology must operate. AI may increase an engineer’s reach, but fundamentals, practical testing and ownership of the final system remain essential.