Building AI that works at the intersection of data and clinical practice, translating medical images, electronic health records, and multi-modal signals into tools that help clinicians act earlier, decide better, and do more for patients.
I am a biomedical engineer with an M.Tech from the National Institute of Technology Rourkela, working at the intersection of machine learning and clinical medicine. My research focuses on building AI systems that are not just statistically robust, but genuinely useful at the point of care: whether that means predicting acute kidney injury before it manifests, estimating portal hypertension without an invasive catheter, or enabling pathologists to analyse collagen structure in tumour tissue without expensive microscopy.
My background spans liver disease, oncological imaging, biosignal analysis, and multimodal data fusion. I work with volumetric CT scans, whole-slide histopathology images, EHR-derived clinical tables, and spectroscopic signals, and with the clinicians, radiologists and hepatologists who use them. I have published across peer-reviewed journals and international conferences, received a Best Paper Award, and hold a Gold Medal for academic achievement.
"Optimizing CNN Models for Multiclass Skin Cancer Classification: A Transfer Learning Approach"
BioMedical Signal & Image Processing Lab
"In-host Modeling of SARS-CoV-2 Infections and Metabolic Implications in Disease Severity"
Dr. Madhuresh Sumit, IIT DelhiCan a model trained on thousands of scans generalise reliably enough to augment, or replace, a specialist's first read?
Deep learning frameworks for analyzing medical images across modalities: CT, histopathology whole-slide images, and cross-modal translations between staining types. Work spans skin cancer classification, pancreatic tumour assessment, and collagen morphology quantification in chronic disease.
What signal already exists in routine clinical data that, if surfaced earlier, could change a patient's outcome?
Building interpretable ML models from routinely collected EHR data to forecast critical clinical events. Focus on acute kidney injury in liver failure, where early warning is the difference between intervention and irreversible organ damage. Emphasis on feature interpretability alongside predictive performance.
How do we design models that can reason across imaging, laboratory values, and clinical history, the way a clinician does?
Designing fusion architectures that integrate CT-derived radiomic features with structured clinical parameters. Current work targets non-invasive estimation of portal hypertension in chronic liver disease, aiming to replace an invasive hepatic venous pressure gradient measurement with a machine learning surrogate.
Non-invasive estimation of portal hypertension by fusing CT radiomic features with laboratory parameters to replace invasive HVPG measurement.
ML model predicting 7-day acute kidney injury risk in liver failure patients from routine EHR data. Demonstrated strong clinical relevance for early intervention.
Deep learning framework translating H&E slides to SHG-equivalent images for non-invasive collagen morphology analysis in pancreatic tumour assessment.
ML-based fusion of near-infrared and fluorescence spectra for non-invasive HLB detection. Combined spectral analysis with vegetation indices.
M.Tech thesis. Optimized CNN with transfer learning on 25,000+ whole-slide images. Achieved accuracy comparable to state-of-the-art with reduced compute.
ML analysis of visual evoked potential signals detecting changes in the visual pathway following coffee consumption across six ML classifiers.
Systematic review comparing 2D modalities vs 3D cone-beam CT for upper airway volume. Identified a research gap for AI-driven estimation approaches.
First-order intensity analysis of stromal collagen in chronic pancreatitis using cross-modality image translation with CNNs. Published in Procedia Computer Science.
Mathematical models of viral dynamics and immune responses in SARS-CoV-2 and influenza. Identified potential antiviral targets and gene therapy candidates.
American Journal of Gastroenterology, 2026. doi:10.14309/ajg.0000000000004031
Journal of Clinical and Experimental Hepatology, 2025;15:102726
Lecture Notes in Networks and Systems, 2026:297–311
Procedia Computer Science, 2025;258:365–373
Communications in Computer and Information Science, 2025:267–277
Manuscript in preparation
Research findings only reach their full impact when they travel beyond academic journals. I have contributed to translating complex biomedical and health science concepts into accessible, engaging content: infographics, written articles, and audio-visual media, reaching a wider public audience through BioXspace-Synergising Stem, an e-learning platform focused on science communication.
BioXspace-Synergising Stem (E-learning Platform)
Managed social media outreach and developed educational content to engage a wider audience in science. Created infographics and articles that made research accessible beyond specialist communities.
BioXspace-Synergising Stem (E-learning Platform)
Gained hands-on experience translating complex scientific concepts into accessible formats. Produced infographics, long-form articles, and audio-visual materials for diverse audiences.
Featured science communication pieces created during my time at BioXspace.
BioXspace · 2022. Visual explainer communicating a biomedical concept for a general audience.
View on Instagram
BioXspace · 2022. Long-form article breaking down a scientific topic for a non-specialist readership.
View on Instagram
BioXspace · 2022. Visual explainer on the applications of synthetic biology for a general audience.
2nd International Conference on AI, Computing Technologies, IoT & Data Analytics
National Institute of Technology Raipur, November 2024
B.Tech Biotechnology
Highest CGPA in entire graduating batch, Banasthali Vidyapith
M.Tech at National Institute of Technology Rourkela
Awarded based on rank in the national GATE examination
All India Rank 573 in Graduate Aptitude Test in Engineering, Biotechnology stream
I am open to research collaborations and conversations about AI in healthcare. Don't hesitate to reach out.