VISWAM.AI

Viswam.ai, is an initiative dedicated to leveraging AI for the empowerment and development of the Global South, with focus on the inclusion of marginalized underserved communities. A collaborative effort between Swecha and IIIT Hyderabad, it envisions creating innovative, inclusive, and accessible AI technologies that can power grassroot solutions. Fostering linguistic, cultural, and social diversity of the regions. Thru cutting-edge research, skilling youth in AI technologies, and foundations to help building solutions for real-world challenges faced by communities. A key priority of Viswam.ai is to build for resource-poor languages/areas, that often lack digital resources & quality datasets required for building AI models. The initiative focuses on developing tools and techniques to compile datasets, and then build effective algorithms with alternatives that reduce computational costs without compromising performance; ensuring that AI solutions are both accessible and sustainable. Starting with AI tools for local languages such as speech recognition and natural language processing models, going on to build purpose specific LLMs. These efforts will contribute to fostering inclusivity in AI. Viswam.ai will also embark on research needed, like developing Bit LLMs and Bit embeddings to enhance efficiency and accessibility for resource-constrained environments. TTS analysis to identify solutions that require minimal GPU resources while providing greater control over output. Efforts in semantic chunking and reducing the size of semantic search models aim to make AI-driven information retrieval faster and more scalable. Supported by creative approaches such as building "Chandamama Kathalu" SLMs, and Speech Recognition systems designed to preserve and promote regional traditions and language accessibility. Foundational datasets and models will be at the core. Like collecting extensive ASR datasets to improve the accuracy and adaptability of speech recognition tools.