How AI is rewiring automotive engineering: Gaurav Kakati, CTO–AI, KPIT
The automotive industry is undergoing one of the most significant transformations in its history. Vehicles are no longer defined primarily by mechanical engineering; they are increasingly becoming complex software platforms where millions of lines of code determine everything from safety and performance to user experience and connectivity.
As software-defined vehicles become mainstream, automotive manufacturers are confronting a new challenge: managing unprecedented software complexity while maintaining the rigorous safety, reliability and compliance standards that the industry demands.
According to Gaurav Kakati, CTO–AI, KPIT, artificial intelligence is emerging as the critical engineering intelligence layer that can help automotive organisations navigate this complexity and accelerate innovation without compromising quality.
In an exclusive interaction with Express Computer, Kakati explains how AI is reshaping automotive software development, validation, testing and engineering decision-making, while laying the foundation for the next generation of smart mobility.
AI becomes the engineering intelligence layer
Modern vehicles are built through the collaboration of multiple engineering teams working across different domains, from infotainment and powertrain systems to advanced driver assistance systems (ADAS), sensors and connectivity platforms.
The challenge is not merely building these systems but ensuring they function seamlessly together.
Kakati explains that AI is helping engineers move beyond traditional automation by providing a system-level understanding of vehicle architecture.
“A modern vehicle comprises thousands of lines of code. Features like the infotainment system, lane assist, or door sensors are all separate software systems, built by different teams and often different suppliers, that must work together flawlessly every single time the ignition is turned.”
He points out that identifying the source of a defect in such an environment can be extraordinarily difficult. “When something goes wrong like a sensor mismatch or a software conflict between the ADAS stack and the powertrain, finding it is like searching for a single faulty connection across an entire city’s electrical grid.”
Rather than functioning as a coding assistant alone, AI is increasingly being deployed as an intelligence layer capable of understanding relationships between requirements, architecture, code and validation processes.
“When a requirement changes, AI flags every system that it can potentially impact before a single line of code is rewritten. When a defect surfaces during testing, AI traces it back through architecture, requirements and code changes to find exactly where the chain broke,” he says.
Transforming design, coding and validation
While AI is creating productivity gains across software development, Kakati believes some of the most significant benefits are emerging in areas traditionally associated with complexity and engineering overhead.
Legacy code remains one of the industry’s biggest challenges. Vehicle programs often inherit software components developed across multiple generations of platforms, making onboarding and knowledge transfer difficult.
AI is helping engineering teams understand and navigate these environments more effectively. “Engineers joining a new program can spend months just understanding what exists before writing a single new line. AI is changing that by making legacy code readable, navigable and transferable in a fraction of the time,” he adds.
Beyond code generation, AI is increasingly helping identify design conflicts before they become development bottlenecks.
For example, if one vehicle requirement mandates over-the-air software updates while another requires offline-only operation in specific markets, AI can identify the contradiction across thousands of requirements and engineering documents before development progresses further. “This removes entire cycles of rework and helps ensure what gets built is aligned with platform constraints and safety requirements from the outset,” Kakati points out.
Making validation smarter
Validation remains one of the most resource-intensive stages of automotive software development.
As vehicles become increasingly connected and software-driven, validating every possible interaction across systems becomes exponentially more difficult.
Kakati notes that modern vehicles can no longer be treated as collections of independent components. “The braking system engages with ADAS, which is interlinked to the powertrain. The powertrain is connected to the cloud. When something goes wrong in validation, the failure is rarely where it appears. It is usually the result of how these systems interact with each other.”
AI is helping engineering teams tackle this challenge through dependency mapping, defect pattern analysis and intelligent prioritisation of test cases.
Rather than rerunning every test following a software change, AI can determine which scenarios carry the highest risk and should therefore receive priority. “Instead of re-running everything uniformly, AI prioritises based on the change being introduced, the safety-criticality and concentration of past defects.”
When failures occur, AI can also accelerate root-cause analysis by correlating logs, traces and diagnostic information across multiple systems simultaneously. “The goal is not to replace rigorous validation. The role of AI is to make validation more intelligent, faster and more scalable.”
Moving beyond isolated AI use cases
According to Kakati, much of the industry’s initial AI adoption has focused on standalone productivity improvements such as code generation, document summarisation and defect analysis.
While valuable, he believes the next phase of AI adoption will be fundamentally different. “The shift will be from isolated tools to connected intelligence, defined by AI that understands the full engineering environment. Not just the code, but the requirements behind it, the architecture it sits within, the safety standards it must meet, and the field data coming back from vehicles already on the road.”
This is particularly important because automotive engineering operates under constraints that differ significantly from many other software industries.
“In general software development, probabilistic outcomes are acceptable. In automotive, that logic breaks down entirely. If an ADAS system detects a pedestrian, the response must be identical every single time, without exception,” he explains.
As a result, the industry’s challenge is not simply deploying AI but ensuring AI understands and operates within deterministic, safety-critical environments.
The rise of intelligence-defined mobility
Looking ahead, Kakati believes the industry is moving beyond software-defined vehicles toward what he describes as intelligence-defined mobility.
In this future, vehicles will continuously learn from operational data, feeding insights directly back into engineering processes.
“Software-defined vehicles will demand continuous engineering, not release-driven cycles. Autonomous systems will require AI that learns from every kilometre of real-world driving data and feeds that learning directly back into development.”
The traditional boundaries between development, validation and deployment will continue to blur as connected vehicles generate increasingly rich streams of operational intelligence.
“As vehicles become connected, the gap between field and factory will close, wherein a sensor anomaly detected across a fleet in one region becomes an engineering input before it becomes a customer complaint.”
For Kakati, the future of automotive engineering is not about simply adding AI to existing workflows. It is about fundamentally reimagining how vehicles are designed, validated and evolved throughout their lifecycle. “The goal is to help the industry make that leap from software-defined vehicles to intelligence-defined mobility, where the vehicle is not just built with AI but continuously shaped by it.”
As automotive software complexity continues to rise, that shift may ultimately determine how quickly manufacturers can innovate while maintaining the safety and reliability that modern mobility demands.
