DeepSeek, the open-source AI mannequin from China, has captured international consideration, and since then, the AI world hasn’t been the identical. Whereas ChatGPT served as a wake-up name for India to construct its personal foundational fashions, DeepSeek appears to be the one that would truly flip that imaginative and prescient into actuality – after greater than two years of anticipation.
Whereas Indian IT stays hesitant about constructing foundational fashions, Nagarro’s CTO Kanchan Ray acknowledged that DeepSeek brings capabilities that might be significantly related for Indian use circumstances.
Identical to most Indian corporations, Ray additionally expressed scepticism about DeepSeek with regards to Indian use circumstances because of its low belief issue. “I don’t have any well-formed opinion but…however the fascinating spin to that is [that] they’ve made it open supply,” Ray instructed AIM.
He added that DeepSeek’s open-source nature is the primary promising signal from the corporate. In line with him, the world will use it, evaluation it, type opinions, after which probably construct some use circumstances round it.
“The accuracy is so good that it’s on par with ChatGPT, Gemini, and Llama. That is very encouraging,” Ray mentioned whereas including that if India can surpass the safety points with DeepSeek, the mannequin’s low price will play a giant position in its adoption.
Indian IT corporations don’t make LLMs, and a very powerful motive is the fee. “Not so many Indian corporations might afford spending triple-digit million {dollars} into making LLMs, however now this (DeepSeek) is fascinating,” Ray mentioned. In line with him, since DeepSeek is anticipated to be skilled with a funds of round $5 million, Indian corporations would possibly lastly have a possibility to enter the sector of constructing foundational AI fashions.
Decide a Mannequin and Construct a Service
Prompted by the announcement of the US’ $500 billion AI initiative, The Stargate Undertaking, and China’s open-source “aspect undertaking” DeepSeek, the Indian IT sector would possibly lastly make an AI mannequin. Regardless of unwavering confidence in Infosys co-founder Nandan Nilekani’s imaginative and prescient of creating India the AI use case capital of the world, the stress and the eager eye from Indians would possibly push service corporations to construct an LLM.
Actually, it isn’t as if they aren’t constructing them. For instance, Infosys mentioned it has constructed 4 small language fashions and is within the means of constructing round 100 AI brokers for its purchasers. This proves there are sufficient funds and capabilities for Infosys to step in and probably spend near $10 million on constructing a foundational mannequin.
The identical is true of TCS, whose CEO just lately mentioned that constructing LLMs from scratch doesn’t supply any actual profit. Ok Krithivasan believes that constructing a basis mannequin for regional languages is sensible for democratising expertise.
CP Gurnani, former CEO of Tech Mahindra, instructed AIM that constructing a foundational mannequin is necessary, and that’s what led him to start out the Undertaking Indus initiative throughout his tenure. In the meantime, Gurnani’s AI startup AIonOS has already mentioned that it’s going to work with DeepSeek’s mannequin and supply it to its purchasers.
Since DeepSeek can also be obtainable as an API and the inference price is sort of 9 instances lower than that of OpenAI, India IT providers would possibly lastly undertake the mannequin for constructing providers for its purchasers.
India has the Expertise
Ray mentioned that together with the fee, the talents are additionally an important issue with regards to constructing LLMs in India. “We’d not have created an LLM from scratch, however we’ve got taken LLMs, fine-tuned them to our consumer’s context,” Ray mentioned, including that this is identical with the Indian IT corporations.
If Indian IT providers can construct fashions for his or her purchasers, they positively have the talents and workforce to construct one from scratch.
Ray added that IT providers are extra involved about offering AI infrastructure as a service. That’s the reason Nagarro additionally has its platform, NIA, which hosts totally different frameworks and LLMs for its purchasers.
Evaluating the metaverse wave and the ChatGPT wave to DeepSeek, Ray mentioned that he’s involved about what is going to occur as nobody talks in regards to the metaverse anymore.
Nagarro strongly believes that 2025 is the yr of agentic AI and has developed its personal generative AI playbook. The agency has applied this AI framework throughout varied industries, together with magnificence, healthcare, and automotive. Whereas the core algorithms for forecasting and decision-making stay related, the fine-tuning is extremely context-sensitive.
So whereas the IT giants from India talk about constructing a foundational mannequin, Ray is optimistic about what DeepSeek has to supply for Indian IT corporations: if not a motivation to construct from scratch, a minimum of an adoption of the mannequin by its purchasers.
The put up Why Nagarro CTO Believes DeepSeek May be Good for Indian IT appeared first on Analytics India Journal.
