CognitiveLab
An Open Source First AI Research Lab Building from India, for the World.
About CognitiveLab
CognitiveLab is an open-source first AI research lab founded in Bangalore in May 2023. Our core mission is to build impactful AI technology from India for the world, with a strong focus on democratizing access and fostering innovation through open collaboration.
We develop cutting-edge models and tools, particularly excelling in multilingual AI for Indic languages (like Ambari and Project Nayana) and creating widely adopted open-source software (like OmniParse and the Indic LLM Leaderboard). We were also selected for the Microsoft for Startups program shortly after our inception.
To sustain our research and open-source contributions, CognitiveLab generates revenue through consulting services, helping startups and established companies build MVPs, production systems, and custom AI solutions.
Our Mission & Approach
CognitiveLab is dedicated to developing state-of-the-art AI models in India that create tangible impact globally. We prioritize open-source development to foster innovation, accelerate progress, and ensure accessibility. We balance our focus between critical multilingual/Indic projects and broadly applicable AI tools like data parsers and educational resources (e.g., AI Engineering Academy).
TL;DR - Our Major Achievements
- NayanaRevolutionary multilingual, multimodal model supporting 22 languages with SoTA OCR capabilities
- Omniparse6,000+ GitHub stars, 10,000+ monthly users, recognized as one of the fastest-growing open-source repositories
- AmbariIndia'ss first bilingual Kannada-English LLM, evaluated by NVIDIA and Microsoft, featured by Meta
- Meta GrantReceived Llama Impact Grant for advancing multilingual AI capabilities
Why CognitiveLab Exists
The Language Gap
Despite India'ss linguistic diversity (22 official languages), AI development historically overlooked many regional languages, creating a digital divide for over 500 million non-English fluent speakers.
CognitiveLab aims to bridge this gap by building high-quality AI for underserved languages, ensuring technological equity.
The Resource Imbalance & Open Source Need
Access to resources and datasets for Indic AI was limited, often controlled by large entities with less focus on open community involvement.
We champion an open-source approach to prove impactful AI can emerge from India with focused engineering and collaboration, empowering local researchers and developers.
How CognitiveLab Has Evolved
From a bootstrapped initiative in May 2023 to securing international grants, our journey reflects our growing impact. Here are some key milestones:
Founding & Microsoft for Startups
CognitiveLab founded as an open-source first research lab in Bangalore. Accepted into the Microsoft for Startups program around the same time.
Ambari Launch
Released India'ss first bilingual Kannada-English LLM (Ambari), achieving SoTA performance with limited resources.
Indic LLM Infrastructure
Launched tools and benchmarks like the Indic LLM Leaderboard to support Indic language AI development.
OmniParse Launch
Released OmniParse, an open-source data parsing tool that quickly gained traction (6,000+ GitHub stars).
Nayana OCR @ NAACL
First paper on Nayana OCR accepted at the prestigious NAACL conference workshop.
Meta Llama Impact Grant
Awarded the grant from Meta (Llama Impact Grant) to advance multilingual AI (Project Nayana). Public announcement on April 29, 2025.
Ongoing Research
Continuing work on Nayana, OmniParse, Indic infrastructure, and exploring new frontiers in open-source AI.
01 Ambari
India'ss first bilingual Kannada-English LLM, set a new benchmark by being SoTA at the time of its launch. Trained with a modest budget of just $1,000 on Azure'ss infrastructure, it showcased how powerful AI can emerge even with limited resources.
References & Resources
02 OmniParse
An open-source tool designed to ingest and parse any type of data into a structured format. With 6,000+ GitHub stars and 10,000+ developers using it monthly, it'ss rapidly gaining traction in the AI space.
References & Resources
03 Indic LLM Infrastructure
We'sve developed several tools to support Indic language AI development, including the India LLM Leaderboard, Indic Eval, and Indic Tokeniser.
References & Resources
04 Project Nayana
A revolutionary multilingual, multimodal, multitask language model that supports 22 languages, including text, audio, and vision capabilities.
References & Resources
05 Llama Impact Grant
In recognition of our work, particularly with Project Nayana and the Indic LLM Leaderboard, CognitiveLab was awarded the prestigious Llama Impact Grant by Meta. This significant support, set to be publicly announced on April 29, 2025, will accelerate our efforts in advancing multilingual and multimodal AI for diverse languages.
How We'sll Utilize the Grant
Advancing Project Nayana and Indic language AI
- Expand language coverage across all 22 official Indian languages
- Deepen multimodal capabilities connecting text, audio, and visual modalities
- Develop high-quality training data for low-resource languages
- Enhance the Indic Tokenizer to better handle morphological richness
- Create inference optimization techniques for deployment on constrained hardware
Why We Choose Open Source
CognitiveLab'ss commitment to open-source AI is deeply rooted in our foundational philosophy of democratizing artificial intelligence. We believe that open source has the most direct and widespread impact, benefiting both developers and end-users.
Accessibility for All
Open-source AI fundamentally aligns with our mission to make advanced AI technologies accessible across diverse communities, especially in regions where language barriers have historically limited technological inclusion.
Democratization of Innovation
By embracing open-source models, we'sre enabling developers, researchers, and organizations throughout India and beyond to build on powerful foundations without prohibitive costs.
Community-Powered Development
The vibrant ecosystem around open-source models has accelerated our progress through collaborative debugging, shared improvements, and collective problem-solving that proprietary approaches simply cannot match.
Indigenous Innovation Focus
Open-source allows us to develop locally relevant AI solutions that address uniquely Indian challenges while contributing to the global AI ecosystem.