Bhavish Aggarwal’s Krutrim has come out of the gates fast and furious in a bid to create ‘AI For India’
If there were any doubts about generative AI being the biggest thing in tech right now, the unicorn round for Krutrim (कृत्रिम) just two days ago, put those to rest.
Bhavish Aggarwal’s AI startup is not only the highest-valued Indian startup in this segment, but it has also come out of the gates fast and furious in a bid to create ‘AI For India’. As the founder with a third unicorn under his belt said, “Krutrim’s innovative AI solutions underscore the confidence investors have in our ability to drive meaningful change out of India for the world.”
Now, as Krutrim prepares for its public debut next month, it’s pertinent to see if the startup can challenge the dominance of OpenAI, Google and Meta in the GenAI space and put India on the global AI map. But that’s after a look at the top stories from our newsroom this week:
- RBI’s Licence Whip: Fintech companies got a rude shock when the RBI cracked down on the payments aggregator ecosystem last year, and the ripple effects are still being felt, particularly for Instamojo, one of the worst impacted startups
- BYJU’S Many Maladies: Nearly two years late but BYJU’S finally files its FY22 financials, and the auditor’s notes throw up several red flags for the ailing edtech giant, besides the massive INR 8,245 Cr in losses
- Moj In Hot Waters: ShareChat’s short video app is gasping for breath, and despite millions burnt, Moj has not exactly brought the desired results. With revenue down and users fleeing the short video platform, can ShareChat save Moj?
Krutrim Grabs The Spotlight
The battle for GenAI supremacy is well and truly on, but as many experts and investors have claimed in the past, thus far, it’s all about Big Tech. For Aggarwal and Krutrim, this is the frontier that needs to be breached.
“India has to build its own AI, and at Krutrim, we are fully committed towards building the country’s first complete AI computing stack.” — Founder Bhavish Aggarwal
As we reported a few weeks ago, Krutrim’s platform is proficient in understanding and processing up to 22 Indian languages with the ability to write and speak in 10 languages. Simply put, Krutrim is a family of large language models (LLM), including Krutrim base and Krutrim Pro which will have multi-modal, larger knowledge and inference capabilities.
The company claims that the model accomplishes better performance on multiple well-known, global, LLM evaluation benchmarks including MMLU, HellaSwag, BBH, PIQA and ARC.
More importantly, Krutrim wants to go a step beyond and aims to develop the entire cloud infrastructure and silicon chips that are critical for AI infrastructure and applications. The startup also has plans to develop indigenous data centres and eventually, server-computing, edge-computing and super-computers.
After Ola and Ola Electric, Krutrim is the biggest bet for Bhavish Aggarwal who is increasingly distancing himself from the ride-hailing business. Ola Cabs appointed Hemant Bakshi as the new CEO of Ola Mobility.
Plus, Ola Electric’s potential listing later this year is another major milestone for Aggarwal after the buzz he created for the EV company on social media last year. But Krutrim is clearly the flavour of the season for the founder.
Building The Foundation
If we look at the layers of gen AI, Krutrim forms the foundational or LLM layer, while infrastructure and tools will be built on top of this layer. The applications that leverage Krutrim’s LLM will take up the top of the pyramid.
The big hope for Aggarwal is that Indian companies — startups and enterprises — would ditch the over-reliance on OpenAI’s GPT tools and Meta’s Llama to adopt Krutrim.
It’s also why Krutrim’s many capabilities are framed in the Indian context. Speaking to Inc42 in December, Aggarwal emphasised that most western AI solutions are not accurate enough for Indian languages and are also prohibitively expensive.
Having joined the GenAI fray a little late, Krutrim will now have to spar with domestic players like Bharat GPT by CoRover.ai, Pragna by Soket Labs, Tech Mahindra-backed Project Indus and Lightspeed-backed Sarvam AI in India, besides the likes of OpenAI, Google, Meta, Elon Musk-backed xAI and others.
Krutrim aims to stand out by developing full-stack AI solutions. The base LLM of Krutrim is trained on an impressive 2 Tn tokens, representing the largest dataset of Indian data used in training to date, according to the company.
Krutrim’s app will be available in beta version for consumers in February 2024. Additionally, it will also be available as an API for enterprises and developers, seeking to create AI applications. The beta version has been available for enterprises and developers since earlier this year, while the more expensive and advanced Krutrim Pro model is anticipated to be available in Q4 FY24 or before March.
So what can enterprises and consumers hope to do with Krutrim?
Ravi Jain, the head of strategy at Ola Electric, who is also looking at product strategy for Krurtrim, told Inc42 earlier that tasks such as interacting with real-world businesses would become easier. He cited examples such as reserving a table at a restaurant, finding recipes for Indian cuisine that are typically hard for conventional search engines and content platforms.
Jain added that Krutrim aims to manage extensive customer support for emerging startups and businesses, while providing real-time information around customer behaviour. The support for Indian languages could even benefit the rural economy significantly, something big tech solutions have ignored thus far.
The model can switch between languages and discuss nuanced topics ranging from poetry in Bengali, to Bollywood movies, to creative recipes, the company said in a statement.
Can Krutrim Catch Up?
One area that Krutrim has not yet addressed is whether it is looking at a generative image or video product in the future. We know the company is looking at generative AI applications with voice access in Indian languages, but generative images have become especially popular for consumer apps.
Krutrim’s multimodal models are primarily geared towards text generation and language interpretation. Image generation models are not in the picture at the moment and will come in due time.
“While we are not there yet, we initially focussed on text and language models. We are actively seeking partnerships within the ecosystem to bring our models to the market and collaborate in creating Indian data sources for these other modalities,” Aggarwal said in December.
The ‘AI For India’ Moment
But Krutrim’s ambition of developing and manufacturing chips in-house optimised for AI compute holds great potential according to one cofounder in the GenAI space. This could be the great big hope for many startups looking at generative AI because hardware costs as they are right now are prohibitive for early-stage startups.
“A made-in-India GPU or AI processor could also be a great resource for startups or smaller enterprises who cannot develop AI products without spending heavily on Nvidia chipsets,” the cofounder added.
In our outlook for generative AI in 2024, we predicted that Indian language models will be a critical pillar of ‘AI for India’ as non-English internet users outnumber English language users in the country.
“The India Stack’s natural progression would be when processes like eKYCs, signing up for schemes, applying to entrance exams, applying for permits/licences could be simplified by LLMs and generative AI. This could be enabled by public-private partnerships where fine-tuned models by private organisations can get data access for specific functions,” All In Capital founder Kushal Bhagia told Inc42 earlier.
Google rolled out support for nine Indian languages for its GenAI chatbot Bard, soon after OpenAI followed suit. India is a big focus area for these AI giants, but for once, an Indian company might have something of a home advantage. And of course, there’s a high ceiling for LLMs made for India from an application point of view, given the relatively inclusive nature of tech in India.
However, the fact is that many companies have built their AI strategy and budgets around OpenAI, Google’s Bard or Llama, and whether the India-specific platform offers enough to compel them to jump ship is yet to be seen. There is also the matter of whether the operations of startups and enterprises are geared towards Indian language users.
Krutrim is set to capitalise on the AI revolution with its deep India focus, but these are still early days, and while it has the funding and the unicorn tag, that itself may not be enough to win the AI race.
Spotlight: Budget Season Brings Expectations
With the Interim Budget set to be announced this coming week, startups have put forward their expectations, even though many startups do not expect major announcements at the budget, given its interim nature.
- FAME And Other Questions: India’s EV industry is brimming with hope for the budget and the biggest expectations are around the extension of the FAME scheme, and lower and standard GST across EV products and services
- Drones For Defence: Drone startups are seeking higher allocation for the development and deployment of drones in the defence sector, besides faster approvals for non-defence applications
- A Budget For Climate Change? Agritech players believe there is a need for developing digital public infrastructure tailored for the sector to mitigate the impact of climate change
Sunday Roundup: Startup Funding, Tech Stocks & More
- Funding Picks Up Pace: Funding momentum across India’s startup ecosystem is gaining some steam — this past week, startups raised $170 Mn, representing week-on-week growth of 25%
- Flipkart’s Momentum: Flipkart is nearing profitability after cutting on its cash burn said CEO Kalyan Krishnamurthy, and it also venture UPI payments to build its fintech empire
- The End Of Broker Network? The NCLT has ordered the initiation of an insolvency process against Rahul Yadav’s company, but does this complicate existing investigations against the startup?
- Zomato’s Payments Run: Listed giant Zomato got the RBI nod to operate as an online payment aggregator as it looks to fortify its payments operations after adding UPI to its cart last year
