Deepayan Basu
B.Tech in Materials Science, IIT Jodhpur
Aspiring Data Scientist with a passion for innovation and problem-solving. Currently pursuing a B.Tech at IIT Jodhpur, I am dedicated to honing my technical skills and contributing to cutting-edge projects.
About Me
I'm currently in my final year of B.Tech in Materials Science and Metallurgical Engineering at IIT Jodhpur. While my academic training lies in materials engineering, I've actively explored the intersection of machine learning, financial analytics, and software development. From radar absorbing materials to financial LLMs and chatbot design, I love working at the crossroads of innovation and impact.
Academic Experience at IIT Jodhpur
I’m currently pursuing my B.Tech in Materials Science and Metallurgical Engineering at IIT Jodhpur. Over the years, I’ve developed strong foundations in physical metallurgy, materials characterization, and computational simulations. I’ve explored interdisciplinary fields like machine learning for material property prediction (band gaps, stress hotspots, etc.), with tools like VASP, LAMMPS, and Python-based data pipelines.
My coursework included: Phase Transformations, Smart Materials, Material Processing, and ML for Materials. These helped me build a well-rounded understanding of both traditional metallurgical concepts and modern computational tools.
B.Tech Project: Band Gap & Stability Prediction in Perovskites
My B.Tech project, under the guidance of Prof. Appala Naidu Gandi at IIT Jodhpur, aimed to accelerate materials discovery using machine learning models that predict electronic properties—specifically the band gap and thermodynamic phase stability—of perovskite compounds.
We extracted over 4,500 perovskite structures from the Materials Project Database using PyMatGen and derived descriptors such as average atomic numbers, ionization potentials, and bond lengths. These were used to train regression and classification models, replacing time-intensive DFT calculations.
For band gap prediction, we used Extra Trees and Random Forest Regressors. For phase stability, we employed both regression (R² = 0.83) and classification (F1 = 0.83), determining whether a material is stable based on its energy above the convex hull (Ehull).
We validated our predictions against actual DFT results and implemented workflows using VASP, VESTA, and VASPKIT for electronic structure calculations. This project demonstrated how AI and domain knowledge can collaboratively speed up the material discovery pipeline.
Internship at DRDO – Defence Laboratory, Jodhpur
During my Summer 2024 internship at DRDO (Ministry of Defence, Govt. of India), I worked in the Camouflage Division to develop advanced Radar Absorbing Materials (RAM) aimed at reducing radar cross-section (RCS) of fighter aircrafts. We used a classified Glass-Carbon Hybrid veil and Epoxy to fabricate samples using the Hand-Layup technique, followed by curing in a Vacuum Bagging system.
I tested the prototypes in an Anechoic Chamber across 8–18 GHz frequency to analyze radiation losses. Further testing included exposure to Rain Test Chamber, Thermal Shock Chamber, and CHNSO Elemental Chamber for environmental durability.
Alongside material research, I developed a joint despeckling–recognition CNN for SAR Target Classification. Using the MSTAR dataset and Task-Driven Pruning, the model achieved 94% accuracy while reducing inference time by 23%. This work bridged radar image processing with efficient deep learning.
Mentorship & Music
Mentor and Core Team Member of Sangam Music Society, IITJ:
Handled multiple instruments and demonstrated proficiency in managing the society’s diverse performances. Oversaw coordination between band members, sound engineers, and venue staff to ensure smooth and impactful shows.
Interests and Aspirations
I’m passionate about AI, NLP, LLMs, and using data science to solve real-world business problems. I recently built FinBot, a chatbot that analyzes financial documents, and I'm learning full-stack development to deploy my solutions at scale. My long-term goal is to bridge the gap between technical research and industry needs through practical, intelligent systems.
