Anuran Bhattacharya

About Me

Hi, I am Anuran! I am an M.Tech. student in Computer Science & Engineering at the University of Calcutta, where I am advised by Dr. Sanjit Kumar Setua. My academic journey is driven by a deep commitment to advancing the frontiers of artificial intelligence. I am passionate about pioneering fundamental and applied research, transforming complex theoretical mathematics into robust, trustworthy computational systems.

Core Research Interests

Computer Vision
Natural Language Processing (NLP)
Physics-based AI
Verifiable Machine Learning
Representation Learning

My research bridges theoretical mathematical foundations with high-stakes, real-world AI deployment. I am particularly passionate about engineering deterministic, physically-grounded algorithmic architectures—such as integrating Partial Differential Equations (PDEs) into neural topologies—and bypassing heavy framework abstractions to build robust, trustworthy systems. For my M.Tech. dissertation, I developed PHORENSICS, a physics-informed computational forensics framework designed for synthetic media detection and Generative AI security.

Previously, I earned my M.Sc. in Computer Science from the University of Calcutta, advised by Dr. Rajat Kumar Pal, where I focused on engineering resource-scarce neural networks for computational linguistics.

Before my master's, I completed my B.Sc. in Computer Science at The Bhawanipur Education Society College. During my undergraduate career, I was fortunate to be mentored by and collaborate with a distinguished group of researchers, including Prof. Sanjib Halder, Prof. Priyanka Banerjee, Dr. Ritajit Majumdar (IBM Research), and Dr. Debasmita Bhoumik (Indian Statistical Institute). These foundational collaborations shaped my research trajectory, enabling me to develop data-efficient machine learning techniques for predictive analytics and highly robust recommendation systems.

View / Download Full CV

Advanced Research & Projects

Axiom Engine: Neuro-Symbolic Mini-LLM

PyTorch Transformers Neuro-Symbolic AI

Engineered a "Tabula Rasa" custom Sequence-to-Sequence (Seq2Seq) Transformer from scratch in PyTorch to parse and resolve complex Ordinary Differential Equations (ODEs). Designed a dual-system neuro-symbolic architecture that routes neural structural pattern recognition directly into a deterministic algebraic kernel (SymPy) for hallucination-free, mathematically verified step-by-step integration.

Check live working here

PHORENSICS: Physics Forensics Engine (M.Tech Dissertation)

Physics-Informed AI PDEs Computer Vision

Engineered a robust, physically-grounded computational framework for Generative AI security and synthetic media detection. Successfully integrated Partial Differential Equations (PDEs) and Global Spectral Analysis to mathematically isolate Generative AI origins far beyond standard semantic AI limits. Published the core framework to PyPI, architecting pure Python routines to guarantee deterministic, reproducible computational outcomes.

AI-Powered Fact Verification System

Python LLMs Sentence Transformers

Developed a hybrid verification framework combining local Gemma3 LLM reasoning, semantic tracking, and iterative self-improvement loops to validate factual claims against dynamically retrieved documents.

TxtVis: Multimodal Latent Space Representation

PyTorch Metric Learning Computer Vision

Engineered a "Tabula Rasa" dual-path PyTorch research pipeline for extracting and visualizing high-dimensional embeddings. Designed a custom Convolutional Siamese Network utilizing Contrastive Loss.

End-to-End Autonomous Vehicle Navigation

Python OpenCV Flask

Engineered an end-to-end behavioral cloning architecture based on the NVIDIA DAVE-2 CNN to learn complex perception-to-control mapping. Developed robust data preprocessing and augmentation engines.

Accommodation Rent Prediction

CatBoost Scikit-Learn

Modeled highly nonlinear relationships within structured real-estate datasets using specialized CatBoost regression, mitigating overfitting through ordered boosting for regional micro-economic variance.

Peer-Reviewed Publications

BEN-RS-ANN: An Innovative Approach for Revealing Emotion from Bengali Text

Authors: Seal, T., Bhattacharya, A., Das, S., Patra, S., Dawn, D. D., Khan, A., & Setua, S. K.

Published in: Applied Computing for Software and Smart Systems (Proceedings of ACSS 2024). Springer LNNS.

A User Independent Recommendation System for Web Series

Authors: Bhattacharya, A., Singhania, A. V., Banerjee, P., Majumdar, R., & Bhoumik, D.

Published in: Emerging Technologies in Data Mining and Information Security (Proceedings of IEMIS 2022). Springer LNNS.