The Reinvention of Product Development in an Era of Intelligent Engineering

The integration of artificial intelligence into product engineering is ushering in a new era of intelligent engineering, changing how manufacturers approach design, simulation, validation and cost optimization. Vijaykumar Ramalingam, Vice President - Engineering, TVS Electronics, shares his perspective on this transition in an interaction with Thiruamuthan, Assistant Editor at Industry Outlook, covering the growing application of AI across engineering workflows and its implications for product development and manufacturing.

Vijaykumar Ramalingam brings nearly three decades of experience in engineering leadership, having worked across automotive, industrial and electronics product development. His career encompasses leadership roles in R&D, mechanical and systems engineering, simulation, digital manufacturing and product innovation. He has also driven engineering transformation initiatives, built multidisciplinary teams and led programs focused on improving product performance, development efficiency and manufacturing competitiveness.

With Generative AI entering engineering workflows, how should product development teams balance AI-generated design recommendations with human engineering judgment and technical accountability?

Generative AI has brought a lot of responsibility, but on the other hand, it has given engineers a wider solution space. Earlier, engineers would look at only one or two design alternatives. This was a time-consuming process. With generative design, engineers can now explore hundreds of possibilities against multiple constraints.

These constraints include cost, weight, thermal performance, reliability and manufacturability. However, generative design outputs cannot be used in the field as they are. While generative design can mathematically provide an optimum design, it may not account for manufacturing tolerances, environmental conditions, component availability and failures that may occur over time.

This means human engineers remain accountable. AI can augment the engineering design process, but engineers are responsible for the final outcome. AI should generate alternatives, identify patterns and recommend design improvements.

The engineering team must handle core activities such as requirements definition, assumption validation and dependency assessment. They must evaluate AI-generated recommendations and approve the design based on manufacturability, customer requirements and other engineering considerations.

For example, a generative design for a PCB may suggest a configuration with a different size or number of components than anticipated. Engineers must validate the design by assessing signal integrity, thermal behavior, EMI/EMC compliance, reliability and supply chain availability before approving it.

Generally, we consider two or three AI-recommended designs, combine their best elements and arrive at a final design based on these parameters. AI can make recommendations, but ultimately, engineers are responsible for validating and approving the final design.

Beyond conventional simulation, how can AI-powered simulation help engineers identify potential product failures earlier in the development cycle?

Simulation has been used in engineering for almost 30 years to validate designs before moving to the next stages of product development.

While conventional simulation validates designs, AI adds learning capabilities by drawing on previous simulations, physical tests, field issues and engineering data. This helps engineers identify potential failures earlier, as resolving issues midway through or towards the end of the product development cycle can be expensive and time-consuming.

For example, in an electronic product, thermal simulation can identify hotspots that may lead to overheating and product failure. AI can further assess combinations of processors, ambient temperatures, enclosure materials and cooling configurations to identify potential overheating issues.

Similarly, in mechanical systems, AI-assisted simulation can analyze stress concentrations, crack propagation, vibration risks and material sensitivity. It can also help engineers evaluate alternative materials, such as aluminum or steel, when conventional materials like copper become expensive.

Another important development is surrogate modeling. AI can learn from previous CFD and nonlinear simulations to create simplified models for what-if analysis. Instead of running a simulation for a week, engineers can obtain results within five to ten minutes. This helps them assess design viability before proceeding with detailed, high-fidelity simulations.

However, AI models require good-quality data, sound assumptions and validated operating conditions. AI will not replace detailed simulations or physical testing, which remain necessary to address uncertainties.

My approach is to fail fast virtually, investigate and validate physically. Potential failures must be identified during the design stage, when changes are less expensive. Once the design is frozen and reaches the tooling stage, making changes can be expensive and time-consuming. This is why AI-assisted simulation is important in product development.

As products integrate hardware, software, and AI, how can systems engineering address the growing complexity of product architecture and integration?

Products are becoming increasingly complex as electronics, mechanical engineering, software, firmware, data and AI come together. It is no longer enough for individual components to work efficiently; the entire system must be dependable and reliable over time.

For example, a smart Point-of-Sale (POS) device depends on its processor, memory, display, printer, scanner, power supply, firmware and application software. A change in one area can affect the overall system. If I change the processor, for instance, it may generate more heat, affecting the product design.

Systems engineering addresses this complexity by translating customer needs into measurable system requirements, allocating them to subsystems, defining interfaces and establishing the overall architecture. It also involves evaluating design trade-offs, component risks and verification plans throughout the product lifecycle.

AI can strengthen this approach by analyzing requirements, identifying inconsistencies, tracing dependencies and generating test scenarios. It can also help identify integration failures, particularly at interfaces between subsystems.

To achieve this, we need digital models and end-to-end requirements-to-test traceability, along with clear ownership across disciplines and well-defined interfaces.

The key shift is from managing individual components independently to managing overall system behavior and interactions. A product is successful only when the complete system works reliably, not just when individual subsystems perform well. Systems engineering provides the structure, while AI helps engineers manage complex dependencies from requirements through validation and testing, including in actual customer environments.

Also Read: How AI and Machine Learning Are Transforming Test Automation

How can AI enable the digital threads to use field performance data to uncover recurring patterns and inform more effective design improvements in subsequent development cycles?

A digital thread connects information across the entire product lifecycle, from requirements, design, simulation, validation and manufacturing to service and field performance. AI makes these connections more valuable by helping engineers identify recurring patterns across different datasets.

For example, a printer may experience recurring print quality issues in certain environments due to paper quality, print head temperature, dust accumulation, duty cycles or recently replaced components. When service records, device logs, manufacturing and design information are connected, AI can identify correlations and help engineers determine the root causes of product failures.

This also helps engineers reproduce failures more quickly and identify suitable solutions, whether through design changes, process improvements or better maintenance practices. However, the learning should not stop at resolving field failures. Insights from field performance must feed back into engineering requirements, design rules, simulations, test plans and component selection. This creates a continuous feedback mechanism for subsequent development cycles.

For this to work, field data must be reliable, failures must be classified consistently, and proposed changes must be validated before implementation. The real value of a digital thread is not simply understanding what happened to a product, but converting that knowledge into a better next-generation product.

Under pressure to improve manufacturing margins, how can AI-driven value engineering identify design changes that reduce product costs without compromising quality?

Value engineering is not simply about cutting costs. It is about delivering the required customer functionality and performance at a lower cost without compromising quality or increasing risk.

AI can help engineers examine the end-to-end cost structure, including component selection, PCB layers, material usage, manufacturing yield, assembly time, energy consumption, warranty and service costs. This helps identify cost drivers and explore alternatives more quickly.

For example, AI-assisted workflows can analyze component databases, optimize PCB layouts, simplify enclosures, reduce assembly steps and identify alternative components. Simulation can then help assess the impact of these changes on thermal performance, EMI, mechanical strength and reliability.

However, every proposed change must be evaluated against the overall total cost of ownership (TCO). A cheaper component may increase field failures, while reducing enclosure thickness may affect durability and increase service costs.

Value engineering is therefore a multi-objective exercise involving cost, performance, quality, reliability, manufacturability and serviceability. Every proposed change must pass the necessary validation and verification gates before implementation.

The best value engineering outcome is not the cheapest product, but one that delivers the required performance reliably at the lowest sustainable total cost.

Also Read: Smart Electronics Manufacturing Powered by Advanced AIDC

As R&D moves towards AI-native engineering, how should manufacturers redesign engineering workflows to make AI integral to product development rather than an add-on to existing processes?

AI-native engineering is not simply about introducing a chatbot or adding an AI tool to existing CAD or simulation workflows. It requires redesigning how engineering data, decisions and validations are integrated into product development.

I see this evolving in three stages: First, digital engineering, where requirements, design, CAD, simulation, PLM and test data are connected. Second, AI-augmented engineering, where AI supports design exploration, simulation setup, test generation and failure analysis. Third, AI-native engineering, where workflows are built around collaboration between engineers, engineering tools and governed AI agents.

For example, when a requirement changes, AI can identify affected subsystems, highlight design dependencies, recommend new simulations and generate draft test procedures. Engineers must validate these recommendations before implementation. This requires trusted engineering data, connected tools, clear governance, cybersecurity and trained engineers who understand AI limitations and domain requirements.

Manufacturers should begin with measurable use cases, establish business benefits and scale gradually rather than adopt a big-bang approach. The goal is to create a learning product development ecosystem where each design cycle builds on previous experience to deliver better products in less time and at an affordable cost.