Artificial intelligence (AI) has undergone a massive transformation over the past decade, evolving from narrow applications in machine vision and predictive modelling to the more complex and versatile world of generative AI (GenAI). The recent popularity of OpenAI’s ChatGPT and the emergence of similar GenAI models since then, has made this space more competitive.
Leading this shift is Nagarro, a global product engineering company that has been at the forefront of AI innovation. In conversations with Anurag Sahay, MD and Head of Data and AI at Nagarro, and Ananda Sengupta, MD, Head of Telecom at the company, we explored how Nagarro is differentiating itself in the competitive AI landscape, tackling challenges, and ensuring responsible AI development.
Evolution into Generative AI
Nagarro embarked on its AI journey in 2016, focusing primarily on machine vision, predictive modelling, and natural language processing (NLP). At the time, AI was largely task-specific, with models trained for singular purposes like object detection or predictive analytics. However, with the advent of GenAI, the paradigm shifted.
“Before GenAI, AI models were trained from scratch for each specific task. Today, we adapt large foundation models to achieve multiple business outcomes, which significantly changes the AI technology stack,” Sahay said.
Unlike traditional AI, which required separate models for different tasks, GenAI enables a single large model to perform multiple functions, increasing efficiency and reducing costs. The shift from “narrow AI” to “foundational AI” means that companies can now leverage fewer models for a wider range of applications.
At its core, Nagarro is a product engineering company, distinguishing itself from traditional IT services firms. “At Nagarro, we emphasise building scalable platforms that integrate AI seamlessly into products,” Sahay explained.
By combining AI with product engineering, Nagarro enhances user experiences, automates software development, and refines decision-making processes within businesses. The company believes in using AI not just as a tool, but as a core element in creating superior technology solutions.
At its core, Nagarro is a product engineering company, distinguishing itself from traditional IT services firms. “At Nagarro, we emphasise building scalable platforms that integrate AI seamlessly into products,” Sahay explained.
By combining AI with product engineering, Nagarro enhances user experiences, automates software development, and refines decision-making processes within businesses. The company believes in using AI not just as a tool, but as a core element in creating superior technology solutions.
The Cost and Efficiency Debate
A major industry concern is the cost of running large AI models. OpenAI, for example, has frequently highlighted the high expenses associated with maintaining its language models. However, efficiency breakthroughs are beginning to change this landscape.
Sengupta points to DeepSeek, an emerging AI company that has achieved a 575% profit-to-cost efficiency ratio. “These advancements prove that it’s possible to build and operate AI models more affordably,” he said, adding that competition from companies in China and other regions will further drive costs down.
One approach to making AI more cost-effective is developing smaller, specialised models that optimise efficiency without compromising performance. Nagarro recognises this trend and works with clients to implement the most practical solutions tailored to their business needs.
Nagarro’s AI expertise extends across various industries, with notable success stories highlighting the real-world impact of their technology. One such example is a sperm motility tester developed using machine vision. This innovation addresses a critical healthcare gap in regions where men are hesitant to seek medical help for fertility issues.
“We built the entire product for the client, incorporating machine vision, GenAI, and synthetic datasets,” said Sengupta. “It’s a perfect example of how AI can be used to solve real-world problems while maintaining user privacy.”
Ensuring Responsible and Ethical AI
With the growing concerns around AI ethics, particularly regarding data privacy and bias, Nagarro has taken a proactive stance on responsible AI development. “For us, responsible AI is not optional,” Sahay emphasises. “We work with enterprises that demand strict data protection and governance, and we’ve been practicing these principles long before GenAI became mainstream.”
Nagarro employs several key strategies to ensure ethical AI deployment:
Localisation: AI models are trained where the data resides, reducing risks associated with data movement.
Guardrails for AI Behavior: Boundaries are set to prevent AI models from engaging in undesirable behaviors.
Explainability & Observability: AI predictions are made transparent and auditable, which is crucial for industries like finance and healthcare.
Despite all this, one of the most debated topics in AI governance is accountability. If an AI system makes an incorrect decision, who bears the responsibility? Is it the technology provider or the enterprise using the system?
“Our goal is to build AI that operates within defined ethical boundaries, but we also assume responsibility alongside our clients,” Sengupta explained. “If something goes wrong, we work together to fix it rather than shifting the blame. Our success is directly tied to the success of our clients.”
What’s Next for Nagarro’s AI Initiatives?
Looking ahead, Nagarro is focused on leveraging AI to enhance its internal processes and increase efficiency in delivering AI-driven solutions. “Our CEO often says, ‘We should be able to eat our own dog food,’” Sengupta shared. “This means not just building AI for clients, but also transforming our own workflows to be more efficient and cost-effective.”
