Research & Engineering Portfolio

Breakthrough Achievement: Developed soil-specific Random Forest calibration achieving 34% accuracy improvement for sandy textures in SAR-based soil moisture retrieval.

Scientific Impact:

• Analyzed 677 paired SAR-soil moisture observations across 5 soil textures
• Discovered counterintuitive vegetation enhancement effect (r=0.743 vegetated vs r=0.380 bare soil)
• Developed information proxy framework predicting calibration success
• Publication: Submitted to Remote Sensing Letters (2024)

Innovation: Combined physics-based understanding with machine learning to solve operational remote sensing challenges.

Context: Satellite-based diagnostic evaluating whether watershed interventions produce a measurable dry-season surface moisture advantage over untreated neighbouring land.

Approach: A local control-ring design — treated points inside each intervention polygon compared against a surrounding untreated ring, with the difference referenced to a pre-intervention baseline. Applied across two successive dry seasons.

What was found: Roughly half the watersheds showed either a persistent or emerging positive dry-season moisture signal. A subset held the advantage consistently across both years; others showed it emerging or weakening. The analysis identifies which watersheds are strongest candidates for detailed field follow-up.

A companion assessment examined pre-monsoon surface moisture behaviour in a single taluk during a drought year, showing that one long-treated watershed retained its moisture advantage while untreated neighbours declined.

Status: Internal diagnostic report (May 2026); IISc Civil Engineering, Bengaluru.

Problem: Satellite passive microwave (SMAP-400m) fails on fewer than 12% of monsoon days over the Harave watershed — precisely when hydrology is most active. Cloud cover blocks the optical downscaling inputs; L-band is attenuated under dense wet canopy.

Approach: Calibrated a dry-season empirical transfer function between Sentinel-1 VV backscatter and SMAP-400m soil moisture, selected from four functional forms using the Bayesian Information Criterion and validated by leave-one-year-out cross-validation.

Key Findings:

• Linear model retained for operational use (yields absolute volumetric estimates)
• Sentinel-2 NDVI did not improve calibration — vegetation structure already embedded in SMAP-400m via VIIRS leaf area index
• Fused record validated against independent rainfall and TDR observations in adjacent Mallaiinupura watershed without recalibration

Output: 129-month continuous monthly soil moisture record, 2016–2026, covering 58 monsoon months where SMAP-400m was unavailable.

Submitted to: International Journal of Remote Sensing (IJRS)

Dataset: Continuous 15-minute measurements of soil moisture, temperature, and electrical conductivity (or dielectric permittivity from 2021) at 5 cm and 50 cm depths, with an additional 15 cm depth from 2021 onwards. Co-located precipitation recorded at the same interval. Delivered as annual CSV files in Indian Standard Time.

Study site: Berambadi agricultural catchment, Mysore Plateau, Karnataka (~84 km²). Part of the Kabini Critical Zone Observatory and the OZCAR network. Mixed rainfed and groundwater-irrigated agriculture under semi-arid monsoon conditions (~800 mm/yr).

Record span: 2016–2025. Covers event, seasonal, and interannual scales including a wet year (2016), drought (2023), and recovery (2024). Sensor transition in 2021 (EC → dielectric permittivity) with one year of overlap to support continuity.

What it provides:

• 944-event rainfall catalogue with sub-hourly resolution
• Depth-dependent thermal and hydraulic response of the soil profile
• One of the few long-term, sub-hourly soil hydrothermal records from the Indian monsoon belt

Submitted to: Scientific Data (June 2026, under review after revision)

System Integration: Coupled 1D crop models (C/C++) with 3D Functional-Structural Plant Models (Fortran/CPlantBox), enabling realistic simulation of soil-plant-atmosphere dynamics.

Technical Achievement:

• Implemented loose-coupling data exchange with timestep synchronization
• Created mechanistic sink-term for AgroC using dynamic root architecture
• Built cross-platform PyQt5 GUI for model configuration
• Packaged workflows with Docker for HPC deployment

Scientific Communication:

• Led monthly project coordination
• Authored review paper on model coupling & digital twins (In Silico Plants, under review)
• Contributed book chapters on LLMs and UAVs in agriculture

Reproducible Science: Created containerized environments for plant modeling frameworks, dramatically improving accessibility and reproducibility.

CPlantBox GUI Docker:

• Integrated VNC viewer for 3D root visualization
• Bundled scientific packages (NumPy, SciPy, VTK)
• DockerHub: satraox/cplantbox-gui

DuMuX-ROSI-Jupyter:

• Pre-configured Jupyter environment for soil-plant simulations
• Coupled multi-phase flow (DuMuX) with root models (ROSI)
• DockerHub: satraox/dumux-rosi-jupyter

Project Website: Developed comprehensive documentation platform for agricultural modeling tools integration.

Features:

• Interactive documentation for crop modeling workflows
• Scientific visualizations and animations
• Hugo-based static site with custom SCSS/JavaScript
• Integration guides for multiple modeling frameworks

Visit Project Website

Thesis Title: “Investigation of signatures of plant roots from non-invasive geo-electrical measurements”

Research Problem: How do plant roots influence soil electrical properties? Can we “see” roots without digging them up?

Created the first coupled hydro-geophysical framework linking root architecture, water uptake, and electrical signatures — enabling non-invasive root phenotyping at scales from centimeters to field plots.

Funding: Belgian FNRS (Fonds National de la Recherche Scientifique) - Grant T.1088.15
Defense: 2020, UCLouvain (during COVID-19 pandemic)
Supervisors: Prof. Mathieu Javaux (UCLouvain), Prof. Frédéric Nguyen (ULiège), Prof. Sarah Garré (Gembloux)

Impact:

• 4 peer-reviewed papers (135+ citations)
• 5 international conference presentations
• Multiple international collaborations (Germany, Austria, Israel)
• Established new research field: Computational Root Geophysics

Research Question: Can we build a mechanistic model incorporating BOTH root architecture and water dynamics?

Innovation — Coupled Framework:

• Integrated R-SWMS (root water uptake) with PyGIMLi (electrical modeling)
• 3D finite element models with 500,000+ tetrahedral elements
• Separated direct (root conductivity) vs indirect (moisture) electrical effects
• First model to achieve this level of physical realism

Key Discovery:
Roots impact petrophysical relations — the standard Archie’s Law doesn’t work when roots are present. We quantified exactly how roots modify the soil’s electrical behavior.

Publication: Vadose Zone Journal (2019), 35 citations
Thesis Chapter: 3
Tools: Python, PyGIMLi, Gmsh, R-SWMS, EIDORS

Research Question: Does electrical anisotropy contain information about root architecture?

Breakthrough Discovery:
Electrical anisotropy is a fingerprint of root organization — the first mechanistic proof that geoelectrical measurements encode 3D structural information.

Methodology:

• Generated synthetic root architectures using C-Rootbox (monocots vs. dicots)
• Computed direction-dependent conductivity tensors
• Extracted geometrical indices (convex hull, depth, width, tortuosity)
• Applied machine learning (PCA + k-NN classification)

Key Results:

• Magnitude component (low frequency): water uptake patterns
• Phase component (high frequency): root architecture directly
• Species discrimination: 95% accuracy using k-NN on electrical signatures alone

Publications: 2 conference papers (Geophysical Research Abstracts)
Thesis Chapter: 4
Collaboration: Prof. Andreas Kemna (Bonn)

Research Question: Can ERT discriminate between plant species in real field conditions?

Field Experimental Campaign:

• Multi-season ERT surveys across 6 grassland species (alfalfa, red clover, chicory, plantain, ryegrass, fescue)
• Controlled water deficit experiment (ForDrought project)
• Integration with TDR sensors for soil moisture validation
• Repeated 3D ERT measurements during drying cycles

Novel Methodology — “Model-Informed ERT Interpretation”:
Run synthetic forward models to test what should happen, then compare to observations.

Major Finding:
Successfully discriminated 5 grass species based on their electrical-hydraulic fingerprints. Species-specific depletion zones were clearly visible and statistically significant.

Publications:

• Plant and Soil (2020), 29 citations
• Field data supported by Région Wallonne (ForDrought D31-1341)

Thesis Chapters: 5 & 6
Collaborations: Prof. Sarah Garré, Dr. Florian Wagner (RWTH Aachen), Dr. Nolwenn Lesparre (Strasbourg)

Comprehensive Review: State-of-the-art in geoelectrical methods for soil-root studies.

Coverage:

• Theoretical background (lossy dielectrics, polarization mechanisms)
• Measured electrical properties of plant tissues (resistive & capacitive)
• Overview of ERT and EIT methods for root investigation
• Petrophysical transfer relations (Archie’s Law, vegetation impact)
• Need for explicit root modeling (limitations of mixing models)

Experimental Work:

• Electrical measurements on root segments (DC resistance, polarization signatures)
• Laboratory characterization of rapeseed root electrical properties
• Time-Domain and Spectral Induced Polarization (TDIP/SIP) experiments

Key Insight:
Existing mixing models don’t capture physics correctly. You need explicit 3D representation of root architecture.

Publication: Vadose Zone Journal (2020), 71 citations
Thesis Chapter: 2
Laboratory Collaborations: Dr. Solomon Ehosioke (ULiège)

Objective: Create reproducible, modular framework for root-soil electrical modeling.

Full Pipeline:

Root Architecture (C-Rootbox)
        ↓
Water Uptake (R-SWMS)
        ↓
3D Mesh Generation (Gmsh + Python)
        ↓
Electrical Forward Model (PyGIMLi)
        ↓
ERT Inversion & Analysis
        ↓
Visualization (ParaView, Matplotlib)

Key Features:

• Automatic mesh generation from root architectures
• Coupled hydro-electrical data exchange
• Scalable from rhizotron to field
• HPC-ready parallelization

Open Science Contribution:

• Shared datasets for benchmarking
• Contributed code to PyGIMLi community
• Trained 5+ students in framework usage

Repository: Available for collaboration (contact for details)

Computational Physics: Extended 2D electrostatic Particle-in-Cell plasma code into fully functional 3D electromagnetic simulation framework.

Achievements:

• Developed Helmholtz coil field generation modules for plasma thruster digital twins
• Implemented MPI/OpenMP parallelization for HPC scaling
• Captured wave-particle interactions and plasma instabilities

Funding: NSF grant ATM0647157
Publications: 3 papers in Physics of Plasmas (69 citations)

Research Project: Numerical modeling of light propagation in semiconductor optical systems using coupled Maxwell-Bloch equations.

Physical System:
Exciton-polaritons in semiconductors — quasi-particles from strong coupling between photons and excitonic resonances, crucial for optical switches and quantum information devices.

Numerical Implementation:

• FORTRAN code development for Maxwell-Bloch solver
• 4th-order Runge-Kutta method for temporal integration
• Finite-difference methods for spatial derivatives
• Adaptive time-stepping for numerical stability

Supervision: Prof. Torsten Meier (Theoretical Physics)
Funding: DFG optical signal processing project

Experimental & Theoretical Physics: Investigated laser-induced photodegradation of dye molecules (Rhodamine 6G).

Laboratory Skills:

• Performed photo-patterning and bio-deposition in cleanroom
• Operated high-value laser systems and AFM with nanometer precision
• Analyzed biomolecule deposition patterns
• Teaching Assistant for Physics 101

Academic Performance: GPA 4.0/4.0

Question: Can freely available satellite vegetation trends serve as an annual spatial indicator of groundwater extraction intensity between CGWB assessment cycles?

Approach:

• Computed the “intensifier fraction” — the fraction of cropland pixels with strongly positive Sentinel-2 peak-NDVI trends (2018–2024)
• Evaluated consistency with CGWB extraction percentages across 6 over-exploited blocks (102–118% of annual recharge)
• Used Arunachala Hill (Tiruvannamalai) as geographic anchor for stratified sampling, controlling for broad-scale climate and topographic gradients
• Ran Sentinel-1 VV detectability analysis to explain why single-pixel SAR trend detection fails where block-level optical detection succeeds

Key Results:

• Intensifier fraction positively associated with extraction percentage: Pearson r = 0.814, p = 0.049, permutation p = 0.061
• Rainfall trends show no evidence of confounding
• Block-level aggregation necessary — per-pixel NDVI changes fall below the single-pixel detection bound

Scope: Proof-of-concept for one district; requires multi-district validation before operational use. Intended as a supplement to well monitoring, not a replacement.

Status: Manuscript prepared; independent work outside institutional affiliation.

Science Communication: Produce educational videos simplifying computational science, data analysis, and modeling.

Content Focus:

• Computational modeling fundamentals
• Data science workflows
• Technical writing best practices
• Research methodologies

Production Skills: Video editing (Final Cut Pro, DaVinci Resolve), 2D animation (Pencil2D), visual storytelling

Watch on YouTube

Rather than treating traffic as an engineering problem, this book approaches it as a complex human system shaped by psychology, cognitive decision-making, and social behavior under stress.

The ideas grew out of daily riding observations in Bengaluru — recurring patterns, predictable mistakes, and the dynamics that make Indian urban traffic what it is.

Published: 20 March 2026
Format: Kindle Edition, 61 pages
ASIN: B0GT7D2NMN

Available on Amazon

Published Monograph: Comprehensive exploration of digital twin technology evolution from aerospace engineering to modern agriculture.

Coverage:

• Historical development (Apollo missions → Industry 4.0)
• Agricultural applications and precision farming
• Integration of IoT, AI, and simulation technologies
• Future directions in digital agriculture

ISBN: 979-8900231457
Available on Amazon

Financial Technology: Developed algorithmic trading strategies and execution systems for clients using Python.

Technical Implementation:

• WebSocket integration for real-time tick data
• Trading strategies based on moving averages
• Integration with broker APIs (Angel Broking, Zerodha)
• Automated stock selection and order execution

Period Context: Maintained active skill-building during COVID-19 pandemic while pursuing academic opportunities.