Independent Research

My independent research and academic collaborations span data systems and AI. This is a collection of separate projects, not a single system or a research agenda limited to context management.

Research directions

Semantic query processing. How can database systems combine relational operations with semantic reasoning while controlling execution cost? Sema explores first-class semantic operators, query optimization, and adaptive execution.

Document retrieval. How can agents navigate large document collections and accumulate the evidence needed for complex questions? DocNavRAG explores document-structured graphs and stateful evidence construction.

Hardware-efficient graph algorithms. How can irregular graph workloads make effective use of heterogeneous accelerators? Tensorized k-TTC search explores portable index construction, community search, and maintenance.

Longer-term exploration

I also explore shared context for multi-agent and human–agent collaboration: how knowledge, memory, and work artifacts can be organized into shared, evolving information environments, while providing the right context for each participant and task. The goal is to let people and agents reuse prior work, carry decisions across tasks, and respond to changing information.

This is a longer-term research direction, rather than a claim that the projects above already implement a complete collaboration system.

• When Intelligence Becomes Abundant: Organizing the Shared Information World of Humans and AI Agents
• The Next Generation of Context Management

The original publications retain their author affiliations. Publications contains the complete research record; Research & Systems separates these collaborations from my institutional system projects.