PhD Researcher · IIT Bombay

Nimatullah

Computational Geophysics Researcher

Solving subsurface mysteries through inversion, deep learning, and high-performance scientific computing.

SCROLL

About Me

Nimatullah 📖
Nimatullah
PhD Researcher in Geophysics
IIT Bombay · India

I am a computational geophysics researcher specializing in inversion, deep learning applications in geophysics, and high-performance scientific computing. My work focuses on solving real-world subsurface problems such as groundwater detection using advanced inversion techniques and machine learning.

I bridge the gap between classical geophysical methods and modern computational approaches, developing efficient algorithms that scale to real-world datasets and deliver actionable insights for resource exploration and environmental studies.

🎓 Education
PhD in Geophysics
Indian Institute of Technology Bombay
· Inversion · Deep Learning
· CGPA - 9.26
Ongoing
MSc in Geophysics
Indian Institute of Technology Bombay
· Geophysics · Numerical Modeling & Data Science
· CGPA - 7.76
BSc in Physics
Jamia Millia Islamia
· Electromagnetism · Mechanics · Thermodynamics · Mathematical Physics
· CGPA - 8.45

Skills & Tools

💻
Programming Languages
Python90%
Julia82%
MATLAB78%
🌍
Geophysics
Gravity Inversion Magnetic Inversion Forward Modeling Subsurface Imaging GSI Field Data
🧠
Machine Learning
Deep Learning U-Net Architecture Data-driven Inversion Neural Networks PyTorch TensorFlow
📊
Visualization
2D/3D Plotting Geospatial Analysis Contour Mapping Matplotlib PyVista
⚙️
Algorithms & HPC
PCG Method MCMC Sampling Parallel Computing Numerical Optimization Sparse Linear Algebra

Projects

Research-driven software tools built for real geophysical challenges

Backend-agnostic Julia Framework for 3D Gravity Inversion
Julia

Developed a high-performance 3D gravity inversion framework using Julia. Implemented Conjugate Gradient (CG) methods to efficiently solve large-scale inverse problems on both synthetic and real datasets.

Julia Inversion HPC PCG Geophysics
🗺️
Basement Depth Estimation using PGNNs and CNNs
Python · TensorFlow

Developed a deep learning framework for gravity inversion to estimate basement depth by comparing Physics-Guided Neural Networks (PGNNs) and data-driven CNNs. Integrated Granser's forward model into a hybrid physics-based loss, improving generalization and ensuring physically consistent predictions.

Python TensorFlow PGNN CNN Geophysics
🗺️
TEM Inversion using Deep Learning
Python

Developed a deep learning-based framework for Transient Electromagnetic (TEM) inversion to estimate subsurface conductivity models from observed data using synthetic forward simulations.

Python Deep Learning TEM Matplotlib SciPy
🧲
Magnetic Inversion using Deep Learning
Python

Developed a U-Net-based deep learning model addressing remanent magnetization and unknown magnetization direction in magnetic inversion.

Python U-Net PyTorch Magnetic Inversion

Research

📄
Publications in Progress
Magnetic Inversion using Deep learning
To solve the traditional problem using deep learning.
In Preparation 2024–2025
High-Performance 3D Gravity Inversion in Julia
Julia-based framework exploiting language-level parallelism for three-dimensional density contrast inversion using iterative solvers with regularization.
Under Review 2023–Present

Field Experience

Hands-on geophysical survey experience with industry-standard instruments across diverse field environments.

📡
TEM Survey 1 TEM Survey 2 TEM Survey 3 TEM Survey 4 TEM Survey 5 TEM Survey 6
TEM
Transient Electromagnetic (TEM)

Used TEM surveys for groundwater exploration and depth sounding. Deployed transmitter loops and receiver coils to measure subsurface conductivity variations as a function of time after current cutoff.

Applications
Groundwater Detection Conductive Layer Mapping Depth Sounding
DC Resistivity 1 DC Resistivity 2 DC Resistivity 3 DC Resistivity 4
DC Resistivity
Direct Current Resistivity and Cross Well ERT

Conducted DC resistivity surveys using Wenner and Schlumberger electrode arrays for vertical electrical sounding (VES) and 2D imaging of subsurface resistivity structures.

Applications
VES Profiling 2D Resistivity Imaging Aquifer Characterization
📻
VLF Survey 1 VLF Survey 2 VLF Survey 3 VLF Survey 4
VLF
Very Low Frequency EM (VLF)

Operated VLF electromagnetic instruments to map shallow conducting structures and geological faults. Used tilt-angle and ellipticity measurements for qualitative interpretation.

Applications
Fault Mapping Shallow EM Survey Structural Geology
🧲
Magnetic Survey 1 Magnetic Survey 2 Magnetic Survey 3 Magnetic Survey 4
Magnetic
Ground Magnetic Survey

Carried out ground magnetic surveys using proton precession and fluxgate magnetometers. Acquired total magnetic intensity (TMI) data for subsurface magnetic anomaly identification and inversion.

Applications
TMI Data Acquisition Magnetic Anomaly Mapping Susceptibility Inversion
Field Data Sources
Geological Survey of India (GSI) datasets
Academic fieldwork at IIT Bombay
Collaborative research surveys

Get in Touch

Interested in collaboration, research discussions, or opportunities? Feel free to reach out.

Contact Information
✉️
Email
24D0455@iitb.ac.in
GitHub
naimat04
LinkedIn
Nimatullah
📍
Location
IIT Bombay, Mumbai, India
Send a Message
✉️

Message sent!

Thank you for reaching out. I'll get back to you soon.

🧲
Magnetic Inversion — Deep Learning Results
U-Net · PyTorch · Remanent Magnetization · Synthetic Benchmarks
0.95
R² Score
5%
RMSE (norm.)
Predicted Susceptibility Slice SYNTHETIC · TEST SET
Magnetic Inversion Result
The U-Net architecture was trained on 40,000 synthetic models with randomised prismatic bodies at varying depths, susceptibility contrasts (0.01–0.15 SI), and Susceptibility. The network directly inverts the observed total-field anomaly to a 2-D susceptibility.
0.95
R² Score
5%
RMSE (norm.)
Training & Validation Loss (MSE) 200 EPOCHS · COSINE ANNEALING
Loss Curve
100
Epochs
32
Batch Size