- Analyzed and curated millions of biomedical data points using data-centric models, generating FAIR data for various downstream tasks.- Developed and implemented information extraction models and pipelines for biological databases (GEO and PubMed), enhancing accuracy by 20%- Fine-tuned a proprietary 3B parameter language model using PEFT techniques, on custom instructions tailored for biomedical data curation tasks in low-resource settings.- Co-authored and published a research paper demonstrating the effectiveness of multi-task learning in significantly reducing the inference time on BioNER tasks by 3x.- Built a GPT-powered tool that can normalize biological terms from user-specific ontologies with 96% accuracy, enabling more accurate and consistent annotation of biomedical data.- Developed Generative AI based POCs using Langchain.

During my MS thesis, I worked on classifying shorter texts into fiction and non-fiction genres building upon the previous work done on document-level classification into fiction and non-fiction genres. We proposed 3 novel syntactic features extracted from parse trees, which included modeling word order variations across genres, quantifying syntactic complexity, and argument-adjunct patterns inspired by the theoretical psycholinguistics literature to get deeper insights into the writing style of such texts. Further, We trained a traditional machine learning classifier using these novel features, achieving an impressive accuracy rate of approximately 92%. Additionally, we fine-tuned a BERT classifier, reaching state-of-the-art performance with an accuracy rate of 96%. Notably, our analysis revealed subtle differences in word order between genres, such as the tendency for verbs to precede adverbs in fiction texts.Our research findings were presented at COLING 2022 in Gyeongju, Korea, demonstrating our contribution to advancing the understanding and classification of textual genres.Supervisors: Dr. Arpit Sharma & Dr. Rajakrishnan Rajkumar

• Designed, implemented, and tested RESTful APIs in Python.• Refactored existing codebase for improved readability and performance optimization.• Developed automated scripts to streamline data processing tasks using Python.

Elected representative of the student body of IISER Bhopal for Computing & Networking Council where I was responsible for organizing various events, talks and workshops for the community. Some of my major contributions during my tenure include introducing competitive coding competitions on in-house platform, facilitating entrepreneurship and tech talks, revitalizing the Photography Club, organizing successful gaming events, workshops, and expanding the council’s reach to engage the entire community.

Assisted in regular teaching activities and programming assignments for the course ECS 102

Conducted research to understand the sentiments conveyed in news articles while examining their writing styles.Led the development of the backend of the fictometer website in Django and the NewsChase app in Python and deployed it on Aws EC2 instance. Mentor:- Dr. Kushal Shah