Bridging Academia & Industry

Translating cutting-edge AI/ML research into production-ready systems that deliver measurable impact across healthcare and industrial sectors. Research focused on medical imaging, predictive maintenance, and causal inference.

Publications

Under Review

Citations

TBD

After Publication

Industrial Impact

$6M+

Annual Savings

Datasets

7K+

MRI Scans Analyzed

Under Review

Multi-Class Brain Tumor Classification and Grade Estimation using Dual-Head Neural Network

IEEE Transactions on Medical ImagingPrimary Author & Researcher2024

Novel Dual-Head Neural Network leveraging EfficientNetB0 (Transfer Learning) for simultaneous multi-class tumor classification (glioma, meningioma, pituitary) and grade estimation, achieving 93.2% accuracy with extreme computational efficiency.

Key Findings

›93.2% classification accuracy on 7,023 MRI scans
›4.0M parameters with 15ms inference time
›89% concordance with radiologist grade assessments
›Threshold-based grading algorithm for precision medicine

Methodology

›EfficientNetB0 with Transfer Learning as backbone
›Dual-head architecture for multi-task learning
›Threshold-based grading using segmented tumor area
›Clinical deployment optimization for low-resource settings

Impact

Enables precision medicine in low-resource healthcare settings with real-time diagnostic capability suitable for deployment in resource-constrained clinical environments.

TensorFlowKerasPythonOpenCVMedical AI

View PDF

In Preparation

A VAE LSTM Architecture for Early Fault Signatures, Time to Failure Projection, and Root Cause Pathway Analysis

IEEE Transactions on Industrial InformaticsPrimary Author & Researcher2024

Novel integration of Variational Autoencoders with LSTM networks for comprehensive predictive maintenance, achieving 48-hour advance fault detection with 92% accuracy and enabling root cause analysis through causal inference.

Key Findings

›92% accuracy in fault detection
›48-hour average warning time before failure
›8% false positive rate vs. 35% baseline
›$2.3M annual savings in industrial deployment

Methodology

›VAE for feature extraction and anomaly detection
›LSTM for temporal pattern recognition
›Causal inference (DoWhy, PCMCI) for root cause identification
›Real-time processing of 10,000+ sensor readings/second

Impact

Successfully deployed in oil & gas facilities achieving 20% downtime reduction. Framework adopted by multiple manufacturing facilities for proactive maintenance strategies.

PyTorchPythonDoWhyPCMCIFastAPIKubernetes

View PDF

In preparation

Causal AI for Industrial Anomaly Detection: Beyond Correlation

NeurIPS 2025 WorkshopPrimary Author & Researcher2025

Explores how causal inference methods can reduce false positives in industrial anomaly detection while providing actionable root cause insights, transforming reactive alerting into proactive analysis.

Key Findings

›60% reduction in false positive rates (75% to 15%)
›70% decrease in mean time to resolution
›85% root cause identification accuracy
›$4.2M annual savings in production environments

Methodology

›Causal graphs from observational data using PCMCI
›DoWhy framework for causal inference and counterfactuals
›Controlled experiments for causal relationship validation
›Automated root cause identification framework

Impact

Paradigm shift from reactive anomaly detection to proactive root cause analysis. Framework validated across 6 industrial use cases saving millions in unnecessary shutdowns.

PythonDoWhyPCMCINext.jsPostgreSQLKafka

Interested in Collaboration?

Open to research collaborations, joint publications, and consulting opportunities in AI/ML, medical imaging, and industrial systems.

Get in Touch→