Do Better Scores Mean Better Physics? What We Learned from Sim2Real Neural Operators
Why a lower neural-operator prediction error can still hide a worse physical forecast, and what that means for evaluating scientific AI.
Shanghai Jiao Tong University · CGS Doctoral Fellow
Somyajit Chakraborty, (Sam) PhD (reading). (叶一明)
Physical AI Researcher · LLMs, CFD & Complex Networks
PhD Scholar, Dept. of Chemical Engineering — Shanghai Jiao Tong University
Research Assistant (Gen AI) — University College Cork
Visiting Researcher — IIT Varanasi (BHU)
Welcome
I am a PhD student at Shanghai Jiao Tong University (SJTU) in the School of Chemical Engineering and Chemistry, where I work on AI for Scientific Simulation, Physical AI, and Computational Modeling under the supervision of Professor Chen Xizhong. My long-term research vision is to develop next generation Physical AI and Intelligent Systems that can accelerate scientific simulation (mainly Computational Fluid Dynamics), surrogate modeling, design, optimization, and discovery in complex physical and industrial systems through machine learning.
Before starting my PhD, I completed my Master’s at University College Cork under the mentorship of Professor Rosane Minghim, a highly respected researcher in visual analytics, data visualization, and scientific data analysis. I also hold a B.Tech. in Computer Science and Engineering from Techno India University. My academic and research journey spans AI, scientific machine learning, computational physics, generative modeling, complex networks, and data-driven discovery.
My current research interests include:
(1) AI for Physical Sciences and Engineering: Building physics-informed and data-driven AI systems for CFD, PDE surrogate modeling, multiscale scientific simulation, process optimization, and scientific computing. The broader goal is to enable more reliable and efficient simulation-assisted engineering.
(2) Optimization in CFD: Building and Improving CFD solvers with improved speed and optimzation using AI, Quantum Computing, GPU assisted HPC. Exploration of CFD-DEM coupling, Finite Element Method, Unsteady Flow, Turbulance Modeling are some of my focus.
(3) Applied Mathematics: Exploring the mathematical foundations of scientific computing and physics-informed learning, with particular interests in computational meshes, fluid mechanics (Eulerian and Lagrangian coupling), differential and geometric topology, manifolds, spectral analysis, Fourier and wavelet methods, graph theory, partial differential equations, inverse problems, numerical optimization, and multiscale dynamical systems.
(4) AI for Material Science and Pharmaceutical Research: Exploring AI-driven automation, micro/nano-robotics for surgical and medical tasks, drug and capsule design using Physical AI, DEM-FEM coupling, pharmaceutical manufacturing, and digital twins for medical and drug-delivery systems.
I am open to research and technical collaborations in Mainland Europe, China, Singapore, and the United States. If you work on AI for Science, physics, scientific discovery, or advanced machine learning for real-world systems, I would be glad to connect.
Research Interests
AI agents & LLMs
Large Language Models, RAG, Multi-Agent Systems
Computational Physics
CFD, PDE Solvers, Physics-Informed ML
Complex Networks
Social Networks, Scientometrics, Cognitive Networks
AI4Science
Scientific Computing, Simulations, Materials Design
Reviewer
-
Journals:
- Engineering Reports
- Ocean Engineering
- ACM Transactions on Computing for Healthcare Conferences:
- ICETCI - The 6th International Conference on Emerging Techniques in Computational Intelligence, 2026
- ICECCME — The 6th International Conference on Electrical, Computer, Communications and Mechatronics Engineering, Bali, Indonesia
- ICECET — 5th International Conference on Electrical, Computer and Energy Technologies, Rome, Italy
My Favorites
- Future: AGE OF BEYOND: If Humans & AI United — This is humanity's future we should all strive for.
- Video: BBC - The Fantastic Mr Feynman — A fascinating BBC feature on Richard Feynman (my all time mentor from the great beyond!), highlighting his genius, curiosity, and unique way of seeing the world.
- Sci-fi show: Star Wars — An epic space opera set in a galaxy far, far away, spanning a rich universe of heroes, villains, and the eternal struggle between light and darkness.
- Sci-fi novel: Isaac Asimov — Foundation series — A sweeping saga of the decline and rebirth of a galactic empire, following a mathematician who uses psychohistory to guide humanity across millennia of history.
Publications
My research contributions spanning AI, Network Science, Healthcare, and Computational Physics
View on Google Scholar
5
Journals
6
Conferences
1
Under Review
3
Preprints
1
Patents
16
Total
Selected Publications
1
Comeback or dropout: study of discontinued researchers at early career stage
2
PIBERT: A Physics-Informed Bi-directional Hybrid Spectral Transformer for Multiscale CFD Surrogate Modeling
3
Trust@ Health: A Trust-Based Multilayered Network for Scalable Healthcare Service Management
4
BioGraphVAE: A Bio-inspired Autoencoder for Structured Data Reconstruction
Journal Articles
Journal
Machine learning assisted inverse design of low resistivity In–Ga–Sn–Zn oxide sputtering targets
Journal
Comeback or dropout: study of discontinued researchers at early career stage
Journal
Influence of Multi-dimensional Social Capital on Structure of Scientific Collaboration Networks: MDSC@ SciCoNet
Journal
Trust@ Health: A Trust-Based Multilayered Network for Scalable Healthcare Service Management
Journal
Enhancing wind energy forecasting in India: A site-specific comparative analysis of machine learning models and Weibull statistics using long-term data
Under Review
Pending
PIBERT: A Physics-Informed Bi-directional Hybrid Spectral Transformer for Multiscale CFD Surrogate Modeling
Conference Proceedings
Conf.
Ml-HCN: A Novel Healthcare System Architecture Based on Multi-layered Network Approach
Conf.
How Can I Grow!!: Analysis of Co-authorship Network of Early Career Stage Researchers
Conf.
BioGraphVAE: A Bio-inspired Autoencoder for Structured Data Reconstruction
Conf.
LeX-Route: Automated Petition Routing in Indian Judiciary with Hybrid Embedding Ensembles
Published
CenterDistill: Weakly-supervised distillation for ambiguity-aware cross-lingual QA
Preprints
Preprint
MedicaNet: Trust based recommender system to recommend doctors with the help of symptoms
Preprint
LLMPR: A Novel LLM-Driven Transfer Learning based Petition Ranking Model
Patents
Patent
基于多智能体协同优化的药物溶出曲线预测方法、设备及存储介质
GitHub Projects
Selected research projects in AI, Computational Physics, Network Science, and Machine Learning
Visit GitHub ProfileEducation
My academic journey across three continents
Current
Doctor of Philosophy
LLMs & Computational Fluid Dynamics
Shanghai Jiao Tong University
Research focus on CFD, DEM, Simulations, LLMs, Physics Informed Machine Learning, and Computational Chemistry.
Honours
Master of Science
Data Science and Analytics
University College Cork
Thesis: "Genchron Network Analytics tool for Medieval Literature Studies" (Honours). Modules included Deep Learning, Machine Learning, Statistical Modelling, Python for Data Science.
Bachelor of Technology
Computer Science & Engineering
Techno India University
Thesis: "Scoring SM Posts using Behavioural Inputs using Social Networks". Grade: 8.27/10
Experience
My professional journey across research, industry, and academia
Current Positions
CGS Doctoral Fellow (PhD)
Shanghai Jiao Tong University
Dept. of Chemical Engineering and Chemistry
Computer Scientist RA (Gen AI)
University College Cork
Dept. of Computer Science
Visiting Lecturer
Techno India University
Dept. of Computer Science and Engineering
Researcher
CAINS Labs
India
Industry Experience
Research
Research Assistant - AI Engineer (LLMs & GenAI)
University College Cork
- Led 3 cross-domain R&D projects leveraging Large Language Models, improving scalability by 40%
- Developed a visual analytics platform for 20+ datasets, boosting insight extraction by 50%
- Designed LLM-based chatbot for real-time path accuracy using PostGIS, up by 15%
- Integrated GCP map plugins for geospatial features, up by 30%
- Evaluated chatbot responses with BLEU, ROUGE (35% improvement)
- Implemented RAG-based approaches for better context awareness (25% improvement)
Industry
Fullstack Software Engineer - Cloud
Virtusa
- Developed 2 SaaS-based cloud apps in ServiceNow environment, reducing manual processing by 40%
- Led team of 8 in project development and system integration, cutting deployment time by 25%
- Integrated payment gateways, MFA, and AWS/GCP services, raising security compliance by 50%
Academic Experience
Academic
Teaching Assistant
University College Cork - Dept. of Computer Science
- Conducted Lectures on Programming with Python I
- Topics: Introduction to Python, Control Structures, Functions, Data Structures, Error Handling, OOP, APIs & Web Scraping, Data Processing, Visualization
- Data Visualization & Machine Learning fundamentals
- Storytelling with Data, Big Data Visualization, Dashboard Design
Academic
Google Developer Club (GDG) Mentor
Techno India University
- Lectures on Machine Learning & Deep Learning with GCP
- Data prep with BigQuery ML, model training, deployment on GCP
- GCP Fundamentals, Compute & Storage, IAM, Cloud Functions, BigQuery, DevOps