India gets into AI mode, plans to build own LLM

31 Jan 2025

India gets into AI mode, plans to build own LLM
Image source: Photo by Ashwini Vaishnaw, licensed under CC BY-ND, via flickr.com
1
Speaking...

The Indian government woke up to reality of its digital status, after a Chinese startup launched a low-cost foundational AI model, and has announced plans to build an Indian version of the Large Language Model (LLM) that would support artificial intelligence applications in India.

India will soon be coming up with its own generative AI model that would rival the latest models of global players like ChatGPT and DeepSeek, union minister for information technology and railways Ashwini Vaishnaw has said.

Addressing the Utkarsh Odisha Conclave on Thursday, the minister said the initiative is backed by the India AI Compute Facility, which has acquired a capacity of more than 18,000 Graphic Processing Units (GPU) that can support Large Language Models (LLMs).

With 18,693 GPUs of computing power in place, India will soon be building its own AI model that would cater to the diverse linguistic, economic, and social requirements of the nation, Vaishnaw said.

The IndiaAI Mission has selected 10 AI startups for building its own model focusing on localised AI applications and enhancing digital accessibility for Indian users. 

The companies include Yotta (which is backed by the Hiranandani Group), Jio Platforms, Tata Communications, E2E Networks, CMS Computers, Ctrls Datacenters, Locuz Enterprise Solutions, NxtGen Datacenter, Orient Technologies, and Vensysco Technologies. Yotta alone will be providing nearly half of the total computing power, with 9,216 GPUs, Vaishnaw said.

The Rs10,370-crore IndiaAI Mission, which is being implemented through seven pillars, including IndiaAI Compute, IndiaAI Innovation Center, IndiaAI Dataset platform, IndiaAI Application development initiatives, IndiaAI FutureSkills, IndiaAI Startup Financing, and Safe and Trusted AI, focuses on localised AI applications and enhancing digital accessibility for Indian users. 

IndiaAI is an independent business division under Digital India Corporation in the ministry of electronics and information technology. 

Vaishnaw also revealed plans to establish AI data centers in Odisha, marking a significant step toward strengthening India’s AI ecosystem and digital infrastructure.

The Indian government has been actively investing in AI research and infrastructure, aiming to reduce dependence on foreign AI models and establish a self-reliant digital ecosystem. The India AI Compute Facility is a critical part of this mission, providing high-performance computing resources for the country’s AI initiatives.

The IndiaAI Innovation Center will support Indian researchers, startups and entrepreneurs to build an Indian Foundational Models. 

Call for proposals

IndiaAI has invited proposals from startups, researchers, and entrepreneurs to collaborate on building foundational AI models trained on Indian datasets. The aim is to establish indigenous AI models that align with global standards while promoting the creation of AI models that suit India’s specific needs.

The models can be Large Multimodal Models or a Large Language Model (LLM) or a Small Language Model (SLM) designed to support specific applications. 

The foundational AI models should be trained on diverse and representative Indian datasets to ensure cultural, linguistic, and contextual relevance. They should also demonstrate capabilities to address India-specific needs in sectors like healthcare, education, agriculture, climate change, and governance.

The LLMs should support creative and scalable use cases with measurable societal and economic impact. They should also have robust mechanisms for bias mitigation, fairness, and ethical AI principles.

The Intellectual Property (IP) rights for the model will remain with the developer, with a provision of a perpetual licence to government for public use.

Proposals will be evaluated on the basis of capability of the team, including its expertise and track record in AI research, development, and deployment.

Multidisciplinary experience in areas such as machine learning, natural language processing, computer vision, and data engineering, besides prior experience in building models as also proven ability to collaborate and deliver high-impact projects, will also be considered in the selection process.

The mission also seeks clarity and innovativeness in the approach to building foundational models, which considers strategy for dataset sourcing, cleaning, and preparation, particularly for under-represented sectors or segments.

It will consider the architecture and methodology proposed for training, optimising, and scaling the model as also the quality, volume, diversity, and interoperability of datasets.

Other considerations include;

Feasibility of using innovative approaches to reduce development time and cost;

A clear plan for the continuous maintenance, evaluation, and enhancement of the foundation models, so as to ensure their adaptability to evolving needs over time;

Aggressive but well-defined timelines for model development; and

Clear milestones for various stages of development with tracking and evaluation.

Selected participant will have to incorporate as a company within 30 days of selection, if not already registered.

For use case development, the defining parameters would be creativity in identifying and addressing societal challenges at scale, feasibility, impact, and scalability of the proposed applications.

Ethical considerations include a commitment to ensuring fairness, inclusivity, and bias mitigation; adherence to data protection, security, and ethical AI guidelines; compliance with DPDP Act, IT Act and all relevant regulatons.

They should adhere to industry benchmarks and best practices in performance, ensure high accuracy, fairness, and robustness, so as to ensure potential for scaling across industries and regions and long-term sustainability.

Financing mechanism

Participants should ensure financial viability and align budget with the proposed funding mechanism (outright funding or venture debt or equity-based investment).

Besides government funding, proposals can be co-financed by industry, VCs, angel investors or non-profit organisations.

Entities already working on models can also apply for Early Stage. Mid Stage or Late Stage development funding. 

They should be cost-effective and justify resources requirement.

Proposals may be funded either directly or through equity contribution, depending on the case.

Initially, there will be direct financial support for model development, subject to milestone-based disbursements. This financial support can be in the form of a direct grant and/or compute credits for AI Compute. This funding can be in addition to the funds raised by the entity through grants from state governments, other ministries or through other sources.

Subsequently, IndiaAI may infuse more funds into the entity through equity on mutual consent and agreement.

Submission of proposals

Proposals must include brief overview of the proposal, detailed profiles of the core team members, approach and methodology for developing the foundational models, data strategy, proposed architecture and training methodology, timeline etc.

They should specify the strategy for continuous improvement to ensure state-of-the-art performance with evolving technology advancements.

Collaborators should provide details of sourcing data, including the need for IndiaAI support; detail potential use cases and their societal impact as well as the feasibility and scalability of applications.

Besides specifying measures to ensure fairness, transparency, and privacy in the model, they should ensure adherence to ethical AI guidelines.

They should provide detailed budget with a breakdown of costs (including personnel, computational resources, datasets, and infrastructure), funding requirements and preferred funding mechanism (outright funding or equity-based investment) besides financial projections and sustainability plan.

Besides, the proposals should include supporting documents, such as relevant publications, datasets, or supplementary materials (if available).

Proposals should be submitted to tenders@indiaai.gov.in with the subject line: Call for Proposals: Foundational AI Models and should reach IndiaAI before 15th of every month.

A panel of experts will review and evaluate the proposals on the bas of the criteria outlined on a first cum first served basis. 

The Call for Proposals window will remain open till a sufficient number of proposals have been selected or for a maximum period of 6 months from the date of the call for proposals.

Shortlisted applicants will be invited for a detailed presentation.

Selected proposals will receive formal funding or investment agreements.

Business History Videos

History of hovercraft Part 3 | Industry study | Business History

History of hovercraft Part 3...

Today I shall talk a bit more about the military plans for ...

By Kiron Kasbekar | Presenter: Kiron Kasbekar

History of hovercraft Part 2 | Industry study | Business History

History of hovercraft Part 2...

In this episode of our history of hovercraft, we shall exam...

By Kiron Kasbekar | Presenter: Kiron Kasbekar

History of Hovercraft Part 1 | Industry study | Business History

History of Hovercraft Part 1...

If you’ve been a James Bond movie fan, you may recall seein...

By Kiron Kasbekar | Presenter: Kiron Kasbekar

History of Trams in India | Industry study | Business History

History of Trams in India | ...

The video I am presenting to you is based on a script writt...

By Aniket Gupta | Presenter: Sheetal Gaikwad

view more
View details about the software product Informachine News Trackers