01
Agentic AI for Scientific Workflows
Training agents to reason, plan, call domain tools, and improve scientific workflows through reinforcement learning and structured feedback.
GRPORLHFRAG
AI × GEOPHYSICS × SCIENTIFIC DISCOVERY
Building intelligent systems for the physical world.
I am working at the intersection of generative AI, reinforcement learning, agentic systems, diffusion models, and subsurface imaging.
ABOUT
Research with industrial impact.
My work combines modern AI with physics-based modeling to solve hard inverse problems in geoscience. I focus on generative modeling, reinforcement learning, multimodal AI, and agentic workflows that can reason over scientific tools, simulations, and domain knowledge.
I have worked across energy technology and applied research, developing methods for seismic imaging, full-waveform inversion, low-frequency extrapolation, scientific foundation models, and AI-assisted subsurface interpretation.
10+Patents
20+Technical publications & papers
15+Years in AI + geoscience R&D
3Core pillars: AI, Physics, Agents
RESEARCH
What I’m exploring now.
02
Diffusion Models for Inverse Problems
Using learned priors, score-based methods, probability-flow ODEs, and posterior guidance to solve seismic and imaging inverse problems.
DPSPF-ODEFlow Matching
03
AI-Augmented FWI & Imaging
Combining wave physics with learned geological priors for velocity model building, full-waveform inversion, LSRTM, and interpretable seismic imaging.
FWILSRTMWave Physics
04
Scientific Foundation Models
Learning reusable representations for seismic data and subsurface models through transformers, self-supervised learning, multimodal learning, and generative pretraining.
ViTDiTSSL
SELECTED PROJECTS
From research idea to working system.
2026
Efficient Seismic Inversion with Diffusion Posterior Sampling
Deterministic posterior sampling with probability-flow ODE solvers for faster, more stable inverse imaging under physical forward operators.
Research →
2026
Agentic RL Training Loop
End-to-end agent training workflows spanning data curation, supervised finetuning, reward design, RL optimization, evaluation, and deployment.
Overview →
2025–2026
Diffusion Priors for Subsurface Model Editing
Geological model generation and editing with diffusion priors for salt-body interpretation, inpainting, structural constraints, and seismic inversion.
Details →
Industrial R&D
Reinforcement Learning for Field Development
Applied deep reinforcement learning to large-scale field development optimization with deployment-oriented constraints and domain simulations.
Case study →PUBLICATIONS & IP
Selected research directions.
Generative AI
Diffusion posterior sampling for seismic inverse problems
Score priors, conditional sampling, PF-ODE solvers, and physics guidance.
Agentic AI
Reinforcement learning for autonomous scientific agents
Policy optimization, reward modeling, tool use, evaluation, and reasoning.
Geophysics
Stochastic tomography and Gaussian-beam depth migration
Inverse methods for sparse seismic acquisition and complex wave propagation.
Machine Learning
Deep learning for seismic acquisition, imaging, and inversion
Attention U-Nets, deep preconditioners, representation learning, and optimization.
TALKS & TEACHING
Sharing ideas across AI and geoscience.
IMAGE 2026
Efficient Seismic Inversion via Diffusion Posterior Sampling
Generative priors, posterior guidance, and probability-flow ODE acceleration.
Industry Research
Inside the Agentic RL Training Loop
How data, reward design, RL algorithms, tools, and evaluation fit together.
Teaching
LLM, RAG & AI Agents
Hands-on teaching and mentoring on modern AI systems and scientific applications.