Adalat AI Partners with Karya for Developing Legal Training Data in Indian Languages

Adalat AI has partnered with Karya to create specialised training data for legal AI models, focusing on Indian accents and languages. This collaboration aims to improve the accuracy of AI in the Indian legal system while providing fair wage opportunities to workers from underserved communities. 

“This partnership will not only enhance the effectiveness of our AI models in the Indian legal context but also create fair wage opportunities for workers from some of India’s most underserved communities,” posted Adalat AI.

Adalat AI acknowledged Karya's early support and mission-driven work, making this partnership particularly significant. The joint effort is expected to have a meaningful impact on both the legal landscape and the lives of those in need.

Founded by Utkarsh Saxena and Arghya Bhattacharya, Adalat AI is incubated at MIT and Oxford with research roots at Harvard, focuses on AI and LLM-driven solutions aimed at resolving case backlogs, delays in the legal system, and the resulting social injustices. 

Collaborating with courts, the company has developed a voice transcription tool tailored for judges and stenographers. Unlike expensive, generic options, this tool is specifically designed to accurately transcribe legal language and Indian pronunciations, ensuring precision in recording witness testimonies, court orders, and judgments.

Karya recently also partnered with Microsoft Research to conduct evaluation of LLMs. The project involved over 90,000 human evaluations of 30 models in 10 Indian languages, completed in under three weeks. This effort is one of the largest multilingual human evaluations of LLMs to date.