Embedded Files
Parameter Efficient Cross-Lingual Transfer | IIT Bombay (Ongoing)
Parameter Efficient Cross-Lingual Transfer | IIT Bombay (Ongoing)
Research done under the guidance of Prof. Preethi Jyothi, IIT Bombay. Working on building a modular and efficient cross-lingual transfer approach for low-resource languages.
Task-Free Multilingual Adaptation: Generalizing Cross-Lingual Transfer Across Tasks | Amazon
Task-Free Multilingual Adaptation: Generalizing Cross-Lingual Transfer Across Tasks | Amazon
Research project done during the 6-month internship at Amazon Alexa. Developed a unified framework to enhance the multilingual capabilities of pretrained large language models without task-specific supervision.
Text Localization for Cross-Lingual Transfer | IIT Bombay & Amazon
Text Localization for Cross-Lingual Transfer | IIT Bombay & Amazon
Collaborating with Prof. Preethi Jyothi (IIT Bombay) and Amazon, this research addresses the scarcity of context-localized data in low-resource languages, like Indian languages. We work towards creating a new benchmark dataset for localized text transfer to Indian locales and creating efficient, fact-based localization methods using large language models (LLMs).
Graph Compression | IIT Kanpur
Graph Compression | IIT Kanpur
Work done in collaboration with Prof. Arnab Bhattacharya(IIT Kanpur), Dr. Oswald C.(IIT Kanpur), Shubhadip Mitra(TCS) and Dr. Arindam Pal (CSIRO). Worked on tackling the problem of graph compression. We developed CGS, a novel unified and efficient compression framework that supports both lossless and user-controlled lossy compression.
Code-mixed Text Translation | IIT Kanpur
Code-mixed Text Translation | IIT Kanpur
For my BTech thesis project, guided by Prof. Umarani Jayaraman (IIITDM), Prof. Arnab Bhattacharya (IIT Kanpur), and Dr. Oswald C. (IIT Kanpur), I developed a Transformer-based translation tool. This tool translates code-mixed English-Malayalam text into English, addressing the linguistic diversity in India. The project tackles the challenges of decoding code-mixed language, including various dialects and the need for a multilingual vocabulary, to improve communication and understanding.
Intention Analysis–driven SMS Spam Detection Using BERT Embeddings | IIT Kanpur
Intention Analysis–driven SMS Spam Detection Using BERT Embeddings | IIT Kanpur
Work done under the guidance of Prof. Arnab Bhattacharya and Dr. Oswald C. from IIT Kanpur. Worked on the problem of efficient spam detection in SMS messages. Proposed SpotSpam, an efficient spam detection algorithm that uses the intention analysis to detect malicious text.
Page updated
Google Sites
Report abuse