I'm a B.Tech CSE student at Heritage Institute of Technology, Kolkata (CGPA: 9.68), working at the intersection of machine learning, deep learning, and real-world data problems. My long-term goals lie in advancing research and innovation at the intersection of AI, machine learning, and healthcare — building systems that make clinical decision-making more accurate, accessible, and equitable. My current work spans academic research at IIT Kharagpur and ISI Kolkata, applied ML virtual internships, and building AI-powered products under hackathon conditions.
- 🔬 Currently researching deep learning for Cyber-Physical Systems security at IIT Kharagpur
- 🧪 Past research at Indian Statistical Institute on ML liver cirrhosis prediction under noise and class imbalance
- 🏆 Participant in StatusCode2 2025, Diversion 2k25 & HACK IITK 2
- 💬 Ask me about ML & DL pipelines, ensemble methods and time-series modelling
- 📍 Based in Kolkata, West Bengal, India
Languages
AI / ML
Frameworks & Tools
Spring Internship 2026 — ISI Kolkata — Liver Cirrhosis prediction Under Noisy and Outlier-rich conditions
Research at IDEAS-TIH, Indian Statistical Institute: benchmarked 16 models on liver cirrhosis clinical data, studying the effect of GMM noise augmentation and class imbalance on ensemble performance. Random Forest achieved F1=0.78 on noisy data — a +0.22 gain over the clean baseline.
XGBoost Random Forest SMOTE Ensemble Methods Clinical ML
Valecta — AI-Driven Hiring Platform
Built at StatusCode2 (IIIT Kalyani @ IISER Kolkata) — a 24-hour MLH hackathon
Led a 5-member team to build an end-to-end AI-assisted recruitment platform featuring resume plagiarism detection (TF-IDF + cosine similarity), adaptive AI interviews, career path prediction, and employer dashboards with automated candidate scoring.
Next.js Flask Python NLP
EV Charging Demand Prediction — AICTE & NASSCOM
Regression model (Random Forest) to predict EV charging demand from 3 years of real-world data. Full ML pipeline: EDA, feature engineering, and model evaluation to optimize infrastructure placement decisions.
Random Forest EDA Regression Feature Engineering
Predictive Maintenance Classifier — IBM SkillsBuild
Multiclass failure-type classifier on 10,000 industrial sensor records using IBM Watson Studio AutoAI. The Batched Tree Ensemble (Snap Random Forest) ranked #1 with 99.5% cross-validated accuracy across 5 failure modes.
IBM Watson Studio AutoAI Random Forest Multi-class Classification