ABHISHEK S

Lead research and development efforts for Text-to-Speech (TTS) models as part of Bhashini’s central government initiative under Digital India. Fine-tuned two major open-source models: AI4Bharat and IITM, enhancing their performance for diverse language processing. Conducted research on voice-cloning techniques to improve the adaptability and personalization of TTS services. Implemented optimization strategies that increased TTS output quality, achieving a 20% improvement in naturalness and intelligibility metrics. Expanded the IITM model by incorporating four additional languages—Sanskrit, Konkani, Maithili, and Nepali—while also updating several existing languages. Additionally, introduced bilingual models for languages such as Kannada, Telugu, Malayalam, Tamil, Sanskrit, and Konkani. Developed a GenAI ChatBot for ADSS Traders and Dashboard, leveraging advanced prompt engineering techniques (CoT) and architecting a real-time streaming response system. Developed a GenAI Bot for Dubai Airports (POC) and a GenAI Legal Bot for Adani Cements, integrating real-time data and advanced search systems (RAGAS metrics).

Developed machine learning models for regression and classification. Performed exploratory data analysis to extract insights and identify relationships within the data. Visualized data by distinguishing between continuous and categorical variables.