RESEARCH LIBRARY

Research

Machine LearningConsumer BehaviorEducation

RESEARCH SUMMARY

My work focuses on building Domain Intelligent Digtial Twins and my independent research interests revolve around AI, consumer behavior and education. I work on physics-aware machine learning at Congruence Systems where we build models that utilize known information about system dynamic operating conditions to accurately reflect real-world system operations. My research examines decisions that influence purchase and adoption habits, investigating why intention so often fails to become action, across aspirational luxury purchases, e-learning adoption and the sustainability gap in fast fashion. I use interpretable machine learning to identify which factors drive these decisions with the intention of providing meaningul data for purpose-driven, resource-constrained businesses.

Publications

[Patent-1]2026

Methods and Architecture for Domain-Adaptive Digital Twin-Based Predictive Monitoring in Physical and Physiological System

M. Nair

Machine Learning
patent ↗

Disclosed are methods and systems for domain-adaptive, digital twin-based predictive monitoring and diagnosis applicable to physical and physiological systems. The system features a physics-aware, dynamics-capturing Extract, Transform, and Load (ETL) pipeline configured to construct operating envelopes, establish steady-state equilibria, detect transient-onset-driven segments of dynamic responses, apply compensatory time-shifts, and generate a homogeneous, column-wise dynamic response data stack. Rather than relying on rigid, predetermined basis functions such as Fourier or wavelet transforms, the system employs subspace decomposition and time-series embedding to dynamically synthesize domain-invariant, endogenous orthonormal temporal bases directly from the dynamic response data stack and isolate corresponding temporal basis coefficients representing underlying system parameters for digital twin training. Furthermore, a closed-loop self-adaptive mechanism continuously monitors reconstruction residuals to autonomously update model parameters and adapt operational boundaries.

[Paper-1]2025

Feature importance analysis using ensemble decision tree models for aspirational luxury purchase intention of young cohorts in emerging market

M. Nair, J. Nair

Machine LearningConsumer Behavior
[Paper-2]2024

The impact of the COVID-19 pandemic on e-learning adoption in an emerging market: a longitudinal study using the UTAUT mode

M. Nair, J. Nair, A. Chellasamy

Machine LearningEducation
# now building
Visit Congruence Systems ↗opens congruencesystems.com
05 · CURRENTLY BUILDING
Congruence Systems logo

Congruence Systems

Congruence Systems is currently building the Dynalitix platform, Domain Intelligent Digital Twins for Industrial and Physiological Systems at scale. At present, we are engaged in filing patents to protect our IP that advances Congruence's vision of physics-aware, dynamics captured machine learning as well as building the MVP for the platform. The platform enables Retrieval Augmented Forecasting of systems that produce sequential or time series signals attuned to identified system operating conditions. Our proprietary IP enables the building of an operating envelope that reflects real-world system operating conditions. Further, we are also developing IP that allows for efficient time-series forecasting of industrial and physiological signals grounded in Koopman Operator Theory.