Madhusikta Tanaya SinghDT

About

I'm a Computer Science (AI & ML) undergraduate at ITER, Siksha 'O' Anusandhan University, Bhubaneswar, currently in my final years and maintaining a CGPA of 8.88. My work centers on NLP and applied machine learning — building systems that hold up to rigorous evaluation, not just impressive demos. Through multiple research internships at IIT Kharagpur, I've built hybrid ensemble models for readability assessment across English and Hindi (Devanagari) text, including a five-model ensemble reaching R² = 0.965, and a three-stage assistive language reconstruction pipeline (FLAN-T5 + generate-and-rerank + lexical guarding) that improved SARI scores by 15 points over baseline. My encrypted network traffic classifier reached 0.9891 AUROC and was published at CyberAI2024, and I presented a multi-task readability ensemble at RAAM 2026, IIT Hyderabad. I currently have several papers under review at Scopus-indexed venues. Beyond readability and NLP, I've built forecasting systems combining LSTM, attention, and Kolmogorov-Arnold Networks for agricultural price prediction, and I've won 6+ national hackathons designing AI products end-to-end — from architecture to pitch. I'm looking toward a funded MS, where I can keep pushing on the intersection of NLP, interpretable ML, and real-world deployment.