Mr. Pratham Sunkad, at IIT Madras, India.
Pratham Sunkad is an aspiring mechanical engineer and AI researcher from the Indian Institute of Technology, Madras (IIT Madras). With a strong foundation in computational mechanics and artificial intelligence, he has worked on cutting-edge projects in Variational Physics-Informed Neural Networks (VPINNs), numerical mathematics, and fluid-structure interaction. His interdisciplinary approach combines machine learning with computational physics to solve real-world problems. He has interned at American Express, the Indian Institute of Science (IISc), and Weierstrass Institute, Berlin, contributing to AI-driven optimization and scientific computing. Passionate about physics, engineering, and artificial intelligence, Pratham is actively involved in research that bridges machine learning with classical numerical methods. He has published in reputed journals and arXiv preprints, making significant contributions to computational science. Beyond academics, he enjoys music, sports, and mentoring underprivileged students for JEE Advanced.
Professional Profile
ORCID
🎓 Education
Pratham is currently pursuing a B.Tech in Mechanical Engineering with a Minor in Artificial Intelligence at IIT Madras (2021–2025), maintaining an impressive CGPA of 8.67/10. He completed his Class XII (96.2%) at National Public School, Rajajinagar, Bangalore, excelling in Mathematics and Computer Science. His coursework spans diverse fields, including heat transfer, fluid mechanics, machine learning, finite element analysis, and computational fluid dynamics.
Throughout his academic journey, he has displayed a strong aptitude for multivariable calculus, probability & statistics, differential equations, and deep learning techniques. His technical expertise includes programming languages such as Python, C++, C, Java, SQL, and MATLAB, along with proficiency in TensorFlow, PyTorch, NumPy, and Scikit-learn. His rigorous academic training, complemented by hands-on research, has prepared him to tackle interdisciplinary engineering and AI challenges.
💼 Experience
Pratham has gained valuable industry and research experience through internships at leading institutions:
- 🔹 Data Analyst Intern, American Express (May–July 2024): Optimized merchant outreach strategies using predictive analytics and machine learning. Developed algorithms to identify high-risk cases for email engagement.
- 🔹 Research Intern, AiREX LAB, IISc Bangalore (Dec 2023–Dec 2024): Worked on Variational Physics-Informed Neural Networks (VPINNs), integrating finite elements with machine learning for scientific computing.
- 🔹 Research Intern, Weierstrass Institute, Berlin (April–Dec 2024): Implemented Streamline-Upwind Petrov-Galerkin (SUPG) methods to enhance VPINNs for convection-dominated problems. Developed an adaptive indicator function to improve accuracy.
These experiences have strengthened his skills in computational data science, numerical mathematics, AI-driven modeling, and scientific computing.
🔬 Research Interests
Pratham’s research focuses on combining machine learning with numerical methods to solve complex engineering problems. His key interests include:
- ⚡ Physics-Informed Neural Networks (PINNs) & VPINNs – Bridging finite element methods with deep learning for better physical modeling.
- 🌊 Computational Fluid Dynamics (CFD) & Fluid-Structure Interaction (FSI) – Using advanced numerical methods like Lattice Boltzmann to model fluid mechanics problems.
- 🧠 Deep Learning & AI for Engineering Applications – Applying deep learning models to optimize structural and fluid simulations.
- 🔍 Optimization & Uncertainty Quantification – Studying how machine learning can improve accuracy in numerical simulations.
His projects on medical image segmentation, wavelet-based PDE solvers, and reinforcement learning further extend his expertise in AI-driven scientific computing.
🏆 Awards & Achievements
Pratham has received multiple national-level accolades, highlighting his academic excellence:
- 🥇 JEE Advanced (2021): AIR 1866 among 1M+ candidates.
- 🥈 JEE Mains (2021): AIR 1548 and 100 percentile in Physics among 1M+ candidates.
- 🥉 KVPY Fellowship (2020): AIR 1227, among the top 0.6% of candidates.
- 🏅 INPhO Finalist: Ranked among the top 200 students in the Indian National Physics Olympiad.
- 🎵 Trinity College London – Grade 6 Piano (Distinction).
- 🥋 Black Belt (Dan 1) in Karate from Kaishogun Karate Do India.
These achievements underscore his proficiency in competitive problem-solving, scientific research, and extracurricular excellence.
📚 Top Noted Publications
Pratham has co-authored high-impact research papers in AI-driven scientific computing.
1. Improving hp-Variational Physics-Informed Neural Networks for Steady-State Convection-Dominated Problems
Authors: Thivin Anandh, Divij Ghose, Himanshu Jain, Pratham Sunkad, Sashikumaar Ganesan, Volker John
This paper proposes two significant enhancements to the FastVPINNs framework for tackling convection-dominated convection-diffusion-reaction problems:
-
SUPG Stabilization Term: Incorporation of a term inspired by Streamline Upwind/Petrov-Galerkin (SUPG) stabilization into the loss functional. The network architecture is designed to predict spatially varying stabilization parameters.
-
Adaptive Indicator Functions: Introduction of a network architecture that learns optimal parameters for a class of indicator functions, improving the enforcement of Dirichlet boundary conditions.
Numerical studies demonstrate that these enhancements lead to more accurate solutions compared to existing methods.
2. FastVPINNs: Tensor-Driven Acceleration of VPINNs for Complex Geometries
Authors: Thivin Anandh, Divij Ghose, Himanshu Jain, Sashikumaar Ganesan
This work introduces FastVPINNs, a tensor-based advancement aimed at reducing computational overhead and improving scalability of hp-VPINNs, especially in complex geometries. Key features include:
-
Optimized Tensor Operations: Utilization of optimized tensor operations to achieve significant reductions in training time.
-
Enhanced Performance: With appropriate hyperparameter selection, FastVPINNs outperform traditional PINNs in both speed and accuracy, particularly for problems with high-frequency solutions.
Conclusion
Pratham Sunkad is a highly promising researcher with strong academic credentials, technical expertise, and diverse research experience. His international collaborations and AI-driven physics research are notable strengths. However, to be a top contender for the Best Researcher Award, he should focus on publishing in top journals, leading independent research, and translating work into patents or industry solutions.