The Startup Building India’s Next Generation of Applied Inte…

When the Ministry of Defence wanted to build India’s first indigenous, air-gapped language models, both large and small, for the Indian Army, it entrusted Neuralix AI with the responsibility. The company had secured the iDEX grant in 2024 after its no-cost pilot with the ministry saw success. 

But, developing LLMs in defence environments presented unique constraints. “This is not like any other language model,” said Neuralix co-founder and CEO Vikram Jayaram. “The military writes its own language, the forms, the jargon, the way they interpret results. It’s very different from regular natural language.”

The team spent months understanding military operational workflows and domain language. They redesigned transformer architectures, built custom tokenisers, and trained models from scratch to reduce hallucinations and improve domain inference.

The result was a secure decision-support system that can operate entirely within air-gapped Army networks.

Jayaram built the Bengaluru-rooted startup after decades in aerospace, academia and the energy sector, along with co-founder Rohit Gangwal, who is also a political strategist and the BJP’s state convenor for the foreign affairs department in Madhya Pradesh. The duo wanted to focus on engineering rather than chasing trends.

In a climate where AI is often treated as hype, Neuralix is deliberately contrarian, focused not on trends but on solving hard, persistent problems across industries that have historically struggled to adopt advanced technology. “We have to utilise AI to its limit, where it provides real automation,” Jayaram insisted, emphasising impact over spectacle. 

From NASA Labs to Indian Industry

Jayaram’s diverse journey to Neuralix includes a master’s and PhD in the US, a NASA fellowship, research at MD Anderson Cancer Centre, and a faculty role at the University of Oklahoma. He later led technology at Pioneer Natural Resources. It was here, he said, that the inspiration for Neuralix took shape.

“In legacy industries, adoption is one of the bigger challenges… It takes years to build the infrastructure and scale it. By the time you solve a problem, either the problem changes or the people change,” he explained.

His solution was to build a modular, backwards-solving architecture, a system where problems are broken down into atomic components, and AI models, data transformations, compute, and visualisations can be assembled like Lego blocks. This approach allowed companies to move from concept to deployment with unprecedented speed.

India’s Indigenous LLM

Neuralix’s product aids army personnel by delivering real-time insights from surveillance, weather data, and operational logs, enhancing decision-making in the field. Key applications include forecasting, equipment diagnostics, situational awareness, and understanding both structured and unstructured military documents.

Neuralix even built its own mini data centre to train these models in-house. “We made an investment on the hardware side… something that could host and train these models and deliver the right inferences,” Jayaram said.

Funding for the defence project follows a strict 50-50 structure. He explained that iDEX’s funding strategy consists of 25 crores from the inventor and the remainder from the Ministry of Defence.

One unusual advantage of iDEX is that the intellectual property remains with the startup, he said, adding that it is licensed to the military under a Minimum Order Quantity (MOQ) system, an uncommon but forward-thinking policy for defence software.

Despite the sophistication of Neuralix’s AI systems, Jayaram emphasised that they are decision-support tools, rather than autonomous black boxes.

“There has to be a human in the loop,” he insisted. “We have a separate testing unit, and only when feedback comes in do we modify the implementation.”

This anchors Neuralix’s work across all sectors, ensuring that automation speeds up expertise rather than replacing it.

Expansion in Energy, Manufacturing 

The first successes came in the energy sector, where clients used Neuralix’s platform to achieve electricity savings, predict equipment life, optimise operations, and collect real-time label field data. “All of that could be done now in a matter of days and weeks, not years,” Jayaram said.

Neuralix enhanced its credibility by winning the Shell Changemakers of Tomorrow 2024 competition, which included a grant and collaboration with Shell on deepwater well sediment challenges. The startup developed a method to convert chaotic sensor data into a structured format, improving machine learning analysis and paving the way for advanced alerting systems.

For Jayaram, the project validated his belief that AI must deliver measurable returns.

“As a technology provider, if you are charging a dollar and the client does not make three dollars out of it, then there’s no point in such tools ever,” he said.

While energy remains its strongest commercial vertical, Neuralix is gaining traction in manufacturing, too, the company’s second-largest clientele.

Here, one of the prominent gaps is the lack of labelled operational data, especially in steel plants, furnaces, and downhole sensor systems. Neuralix is helping manufacturers build structured, meaningful datasets by using AI to characterise operational windows, detect anomalies, and convert SCADA time-series data into human-readable “language-like” events.

“Imagine automatically labelling operations… that becomes a language to be utilised,” Jayaram explained. “It still requires several years of building this label set, but natural-language-style queries and chatbots already make a big difference.”

The goal is to help manufacturing plants move from reactive maintenance to predictive intelligence, backed by cleaner, richer operational datasets.

Grounded Innovation, Not Hype

Jayaram is critical of how the world often treats AI as a trend rather than a discipline. “Today, language models are a vogue… tomorrow, quantum will be a vogue. It should be the opposite, what are we trying to solve?” he said. “If we still look at it as a foundational engineering problem, we will use the right technology.”

For Neuralix, India’s AI future must be rooted in problem-solving, measurable ROI, and long-term engineering thinking. And that vision extends to India’s academic ecosystem, too.

“Teach people how to solve problems first. Let them be problem solvers. Then bring the technology,” Jayaram said.

The Way Ahead

With a growing footprint across energy, manufacturing and government services, and a rising strategic role in defence, Neuralix AI is now entering a phase where its architecture can scale far beyond individual deployments. 

The company is preparing for broader rollouts across India’s industrial landscape, while also exploring global avenues where its domain-driven, backwards-solving approach can offer immediate value.

“We have a very mixed clientele at this point,” Jayaram said. 

Neuralix AI has crafted something unusual in the crowded world of AI: a distinctly Indian engineering philosophy, practical, purpose-built and resilient in the face of real-world complexity. 

Its trajectory indicates a future in which India’s AI ecosystem is shaped by problem solvers who comprehend both the science and the stakes, creating technology that endures long after the buzz has faded.

The post The Startup Building India’s Next Generation of Applied Intelligence appeared first on Analytics India Magazine.

Generated by RSStT. The copyright belongs to the original author.

Source