Infosys developing small language model for specific industry applications

Representative image

Infosys is developing its small language model (SLM) using open-source components and its proprietary data sets. Strong focus on Generative AI (Gen AI) has prompted the IT major to work on its SLM, which is being designed for specific industry applications, Chief Executive Officer Salil Parekh said at the company's Q2 earnings press conference.

Parekh said the SLM will be rolled out across various sectors with a narrow data set for specific business needs.

An SLM is an artificial intelligence (AI) model designed to process and generate text, but it is smaller in size and tailored for specific tasks or industries, often using less data and computational resources compared to larger models.

"We are deepening our work in generative AI. We are deploying enterprise generative AI platforms, building our own small language model, and developing multi-agent solutions for our clients," Parekh said.

Infosys' nearest rival and India's largest software exporter, Tata Consultancy Services (TCS) recently said that it is currently managing over 600 AI and Gen AI projects, which are either in production or in the development phase.

TCS, Infosys, and other IT companies have not shared revenue originating from Gen AI deals during the quarter. In the first quarter, TCS reported a pipeline valued at $1.5 billion for AI and Gen AI initiatives.

Earlier, the Bengaluru-based company had said it is working on 200 Gen AI projects, and as many as six out of eight Infosys employees had generated three million lines of codes using large language models (LLMs) in Gen AI, in fiscal 2023-24.

Infosys is designing this SLM for scalability across industries, to help clients build business logic on top of the model for specific use cases. "We are working on different industry applications for the small language model, which will be a huge way for clients to leverage what they can do, building some business logic on top of it," he explained.

Parekh also highlighted the deployment of multi-agent solutions, where agents autonomously handle specific business processes. "We’ve launched multi-agent solutions, beyond just being assistants. These agents work on specific business processes almost completely on their own," he said.

For example, over 70,000 employees of an Infosys client in the telecommunications sector are using the enterprise Gen AI platform to create custom use cases for areas such as customer service, knowledge management, and coding. "We have rolled out this platform, and it’s being used by employees to build their own use cases and see real benefits, not just proofs-of-concept (PoCs) but actual projects that are delivering results," Parekh said.

Also read: Mixed bag for TCS, Infosys, HCLTech, Wipro in Q2: BFSI sees uptick, large deals contract

When asked about the recruitment of talent with Gen AI skills, Parekh said Infosys is “continuing to recruit talent in these areas to expand our capabilities”.

On October 17, the company reported its Q2FY25 earnings and net profit increased 2.2 percent quarter-on-quarter to Rs 6,506 crore from Rs 6,368 crore, missing Street expectations. The Indian IT major’s July-September revenue saw a growth of 4.2 percent on-quarter, reaching Rs 40,986 crore.

A Moneycontrol poll of 11 brokerages had estimated Infosys' fiscal second-quarter net profit to rise to Rs 6,769 crore, and revenue to increase to Rs 40,857 crore. Growth was projected to be driven by the ramp-up of large deals, increasing traction in Generative AI deals, and cost optimisations.