Machine Learning Engineer | Physics-Informed ML, Forecasting, Anomaly Detection, Retrieval Systems
Building machine learning systems for science, simulation and reasoning.
- Sole-author paper on bounded latent degradation dynamics, posted as an SSRN preprint
- Building real-time anomaly detection and document intelligence systems at NTT DATA's AI Center of Excellence
- Research interests: physics-informed neural networks, time-series forecasting, anomaly detection, knowledge systems
Learning Bounded Latent Degradation Dynamics for Stable Rollout and Remaining Useful Life Prediction Sole author. SSRN preprint (not peer-reviewed). Bounded/unbounded latent decoupling for degradation forecasting. 53% improvement over best baseline on Lorenz-63 chaotic system, 5.7% blowup rate vs 100% for all baselines. Evaluated on 5 real datasets: NASA C-MAPSS turbofan, PHM milling, IMS bearings, NASA batteries, Beijing air quality. SSRN Preprint
Neural Simulation of Quantum Interactions in a Confined System ResearchGate preprint. Physics-informed neural network embedding Schrodinger's equation directly into the network architecture. 99.9% agreement with analytical benchmarks. ResearchGate
Logos-SIE: Synthetic Information Ecosystem for Truth Discovery and Retrieval Co-authored technical whitepaper, TwinSimLabs. Large-scale synthetic benchmark modeling the lifecycle of information formation for controlled experimentation in retrieval and trust evaluation. GitHub
| Project | What it does |
|---|---|
| Bounded Latent Degradation Dynamics | Degradation modeling and RUL prediction across 5 real-world datasets (SSRN preprint 7180558) |
| Latent Decoupling Framework | Core bounded/unbounded latent decoupling method with Lorenz-63 chaotic system benchmark |
| TinyEarth | Controlled benchmark comparing S4D, Transformer, ConvLSTM, and Mamba for Earth-surface forecasting on EarthNet2021 (958 test sequences) |
| Quantum-Simulator | PINN for quantum particle simulation via embedded Schrodinger equation |
| Logos-SIE | Synthetic information ecosystem benchmark for truth discovery and retrieval |
| Softmatter-State-Detection | Physics-informed ML for soft matter phase classification from molecular dynamics simulations |
| Linux-Kernel-Driver-Evaluation-System | Kernel module + Python evaluation pipeline (C, systems programming) |
| Domain | Tools |
|---|---|
| ML / DL | PyTorch, TensorFlow, JAX, scikit-learn |
| GPU / HPC | CUDA, RAPIDS (cuML, CuPy, cuDF) |
| Scientific | NumPy, SciPy, networkx, LAMMPS |
| Vision | OpenCV, MediaPipe, YOLO |
| Languages | Python, Go, C/C++, R |
| Infrastructure | Docker, Azure, Databricks, Linux |
- Associate Engineer, NTT DATA AI Center of Excellence (promoted from intern)
- BE in Computer Science and Business, Thapar Institute of Engineering and Technology (2022--2026)


