Abhishek Upperwal is the Founder of Soket AI, a frontier AI research lab focused on ethical Artificial General Intelligence. He holds a Master’s in Computational Data Science from IISc Bangalore, specializing in High-Performance Computing, Distributed Systems, and Large-Scale Model Training. His expertise spans CUDA Optimizations, MPI, and Graph Algorithms, with a strong foundation in scaling AI models efficiently.
At Soket AI, Abhishek led the development of Pragna-1B, a 1.25 billion parameter foundation open source model built from scratch for Hindi, English, Gujarati, and Bengali, featuring one of the most efficient Indic tokenizers. He also created the Bhasha Dataset Corpus, a high-quality, open-source dataset for Indian languages.
Abhishek is also spearheading Project EKΛ, an initiative to build state-of-the-art AI models that prioritize Indian languages while aiming to surpass global benchmarks. EKΛ is designed to strengthen India’s AI sovereignty, ensuring the country not only preserves its linguistic diversity in AI but also leads in cutting-edge model development and AI research.
