Research Intern

Supervisor: Dr. Sruti Das Choudhury

- Spearheaded an explainable AI + data-storytelling clustering pipeline across precision agriculture and pediatric healthcare—grouping 22 Indian crop types using 7 agro-climatic/soil features and segmenting a 500-record hospital cohort—showing that z-score rescaling + removing binary gender prevents charge-dominated clusters and surfaces clinically meaningful cohorts (LOS up to 29 days; charges up to 34,644) for decision support.
- Developed a temporal-embedding visual analytics system for 42 plants from 9 genotypes over 25 days, engineering multi-scale phenotype descriptors (growth rates/accelerations, fourier spectra, wavelet energies, distributional stats) and achieving genotype-aligned DTW clustering (ARI 0.30; NMI 0.62) with cross-validated early-prediction curves and SHAP/LIME-linked causal graphs to explain when/why genotypes diverge.
- Implemented an interactive hyperspectral analysis tool, HyperProbe for calibrated datacubes spanning 517-1700 nm (B=243 bands), enabling rapid pixel/ROI annotation, band-difference + Otsu segmentation (IoU/F1 evaluation), and full-scene classification via 3 model families (MLP/logistic regression/random forest) with built-in ablations that log clicks/ROIs under fixed 5-min label budgets to quantify accuracy-per-effort.

Research Scholar

Supervisors: Dr. Arup Kumar Chattopadhyay, Prof. Amit Kumar Das, Prof. Amlan Chakrabarti

- Engineered FHFAM (FH-FAM), a fuzzy-hypergraph feature selection algorithm, achieving the best mean accuracy (81.43%) and best mean feature reduction (89.28%) across 15 agriculture/remote-sensing datasets (5/15 wins) with 11.08s average runtime and statistically significant accuracy gains over key baselines (Wilcoxon p < 0.05).
- Proposed SIFHFAM, a stage-wise intuitionistic-fuzzy hypergraph selector with a monotone submodular coverage objective and greedy (1-1/e) guarantee, delivering the top average accuracy (≈84%) while pruning ≈99% features (typically retaining < 2%) across 14 high-dimensional benchmarks in ~0.1s/run under 10× repeated 75/25 train-test splits.

Undergraduate Student Research Lead; Senior Research Advisor

Leading and managing student research work in the GenAI Centre of Excellence, guiding projects under academic prerequisites and external interests.

Led GenAI CoE's end-to-end research execution and operations—recruited and onboarded members via interviews, mentored and staffed project teams, coordinated 10+ journal groups, maintained the CoE website, and launched ReelBook (Pearson collaboration) and Medium publishing to scale institute-wide research output and AI upskilling at IEM.

Project Intern

Built MemeMetric, an end-to-end cluster-based cryptocurrency forecasting system by architecting the full data/ML pipeline with automated reporting, and integrated real-time Twitter/Telegram/Reddit sentiment signals via NLP to improve robustness and reduce forecast error/volatility.

Undergraduate Research Assistant

Co-authored an IEM-HEALS 2024 accepted study analyzing Jul 2019–Dec 2022 price dynamics of 20 pharma stocks using multivariate regression, volatility modeling, and event-study methods, and engineered TraderBot, a Flask+MongoDB real-time trading simulator wired to Yahoo Finance for live strategy backtesting and portfolio experiments.

Study Abroad Program

Studied fundamentals of Artificial Intelligence, IoT, Machine Learning & Data Analytics, lectured by Dr. Peter Leong, Dr. Eric Cambria, Dr. Matthew Chua, Dr. Yiliang Zhao, Dr. Gabor Benedek, Dr. Tan Kian Hua and others at NUS.