Akash S

Reasearch Intern

IIT Kharagpur • Part time

Nov 2025 - Jan 2026 • 2 mos

Implemented and compared text simplification techniques (SIMSUM, ACCESS, and zero-shot LLMs like Mistral) on legal judgment datasets using Transformer models. Fine-tuned and evaluated BART-based models with standard readability metrics (GFI, FKGL, FRE) and NLP metrics (ROUGE, BERTScore, SARI). Built reproducible training, inference, and evaluation pipelines in Python using PyTorch and Hugging Face, with results documented on GitHub.

AI/ML Intern

DRDO Hyderabad • Part time

May 2025 - Aug 2025 • 3 mos

Worked on transformer-based deep learning models for defense applications. Built and optimized AI/ML solutions using Python and modern ML frameworks, with a focus on self-attention and encoder–decoder architectures. Collaborated with research scientists on model development, experimentation, and data analysis.