Listing Thumbnail

    Indus : GenAI powered LLM

     Info
    Indus LLM brings the power of generative Al based LLM model. The solution is deployed on AWS and leverages AWS cloud services to make it a comprehensive LLM model. The solution intends to make India based foundational Mode for Hindi and its 37 dialects. The model is intended to cater 650 MN Hindi speaking population.

    Overview

    Tech Mahindra’s Indus GenAI solution empowers all Indic languages that have originated out of the great Indus Civilization. Tech Mahindra, with use of AWS powerful platform and services, have sought to build an Open-Source Large Language AI model to serve the needs of 25% of the world's population.

    The Indus Project uses AWS S3 and EC2 instances and intends to make India based Foundational Model in Indic languages. In the current version Indus LLM supports Hindi and its 37 dialects.

    AWS-Powered LLM model

    The solution's robust architecture is deployed on AWS cloud platform and uses S3 and EC2 services for allowing full flexibility to run the solution.

    1. High performance GPU based instance: The model uses AWS EC2 G5 instances for graphic intensive application and ML inference.
    2. User Management and Content Delivery: Utilizes Amazon Cognito for secure authentication
    3. Storage Powerhouse: Application also uses Amazon S3 for scalable content storage
    4. Database Services: Solution uses RDS which makes it easier to setup, operate and scale relational database.

    Key Features:

    1. Built Indus LLM from scratch which is 1.2 Bn parameter model in Hindi and its 37 dialects
    2. Built from 22 Bn tokens using cleaned Hindi and Dialects data.
    3. Most economical, sustainable and ethical model
    4. No spurious token has only Hindi and its dialects.
    5. Features above Llama 2.0 model on Indic LLM Leaderboard for abstract reasoning.
    6. Indus was deployed on AWS Sagemaker and tested for Inference to a decent performance.
    7. NLP Toolkit- This toolkit is used for cleaning of any data other than Hindi and also transliterates dates and English text. This also removes bad words
    8. Bias Removal Toolkit- We have annotated data created for bias and non bias. A cosine difference mechanism is applied as a classifier to find and remove biased sentences
    9. Bias Removal Dashboard- Dashboard to represent biases removed

    Benefits:

    1. Helps to bridge digital gap for Hindi speaking masses of rural India Enables Indic Digital Transformation
    2. Can be finetuned for specific use cases to cater to the needs of native Hindi users
    3. Enables Digital transformation in Indic languages

    Highlights

    • India’s first grounds up LLM focused on Hindi and its 37 Dialects

    Details

    Delivery method

    Deployed on AWS

    Unlock automation with AI agent solutions

    Fast-track AI initiatives with agents, tools, and solutions from AWS Partners.
    AI Agents

    Pricing

    Custom pricing options

    Pricing is based on your specific requirements and eligibility. To get a custom quote for your needs, request a private offer.

    How can we make this page better?

    We'd like to hear your feedback and ideas on how to improve this page.
    We'd like to hear your feedback and ideas on how to improve this page.

    Legal

    Content disclaimer

    Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.

    Support

    Vendor support

    Please reach out to DL (AI.alliances@techmahindra.com ) for further communication