Multimodal Adaptation
Developing efficient ways to adapt multimodal models to new domains, tasks, and operating constraints.
AI researcher · Multimodal intelligence
Intelligence isn’t static. It adapts as the world evolves.
I am an Applied Scientist at Openstream AI, working across multimodal learning, document understanding, and agentic AI. My research connects visual evidence, language, and structured knowledge to help models adapt to new tasks and produce answers grounded in evidence.
Research focus
Building intelligence that can adapt, reason, and earn trust.
Multimodal Reasoning
Connecting visual evidence, language, and structured knowledge for grounded document understanding and reasoning across related forms.
Trustworthy AI Agents
Building agents that learn from interaction, adapt over time, and make verifiable decisions in real workflows.
Recent research
All publications →2026
OntoForm: Ontology-Grounded Semantic Data Generation for Enterprise Document Understanding
EMNLP 2026 · Industry Track
An ontology-driven document compiler that turns one validated business entity into linked enterprise forms with exact labels for extraction, grounded question answering, and cross-form reasoning.