Research Work

A Novel Hybrid ML/DL Framework for Assessing Dark Oxygen Flux Production in the Clarion-Clipperton Zone

The Clarion-Clipperton Zone (CCZ), a vast stretch of the central Pacific Ocean, is the world's most targeted region for commercial deep-sea mining due to its rich deposits of polymetallic nodules. In 2023, a landmark discovery revealed that these nodules are also the source of a previously unknown biological phenomenon: the production of oxygen at depths far below the photic zone, in complete darkness. This "dark oxygen" may have profound implications for the origins of aerobic life on Earth.

This paper presents a novel hybrid machine learning and deep learning framework to quantify dark oxygen production (DOP) flux across the CCZ and model the potential consequences of large-scale seabed mining on this newly identified ecosystem. By training on available geochemical and hydrodynamic datasets, the framework generates spatially resolved predictions of oxygen flux and risk zones, offering a critical decision-support tool for international seabed mining policy.

Co-authors: Simar Singh Rayat, Susheela Dahiya

A Novel Physics-Informed Deep Learning Framework for Tidal Energy Site Selection

Optimal site selection is one of the most consequential decisions in tidal energy development, determining not only energy yield, but environmental impact and long-term viability. Traditional selection methods rely on empirical surveys and simplified hydrodynamic models that often fail to capture complex, nonlinear interactions between tidal flows, seabed bathymetry, and turbine placement.

This paper introduces a physics-informed deep learning algorithm that embeds hydrodynamic governing equations directly into the neural network architecture, constraining predictions to be physically plausible even in data-sparse regions. The result is a site selection framework that outperforms purely data-driven approaches on accuracy and interpretability, providing actionable, spatially explicit recommendations for offshore tidal energy developers.

Electrogenic Biotransformation of Organic Substrates

Access to reliable, affordable energy remains one of the most persistent barriers to development in low-resource communities. This project investigates microbial fuel cells (MFCs) as a decentralized, low-cost alternative, using the electrochemical activity of naturally occurring soil microbes, fed by organic compost, to generate usable electricity.

Over 18 months of independent research, I designed and fabricated multiple MFC prototypes, systematically varied compost age, container size, and electrode material, size, and spacing, and conducted controlled electrochemical performance experiments. The best configuration (12–18-month compost, a 1 L container, and carbon cloth electrodes spaced 1 cm apart) averaged 1.08 W. Key findings demonstrated consistent voltage output from compost-based bioanodic chambers, with substrate longevity as the primary variable in energy efficiency optimization. This work earned the EPA Award and ISEF qualification, and directly informed the LYKA Lamp project.