Semantic interoperability
How independently designed agents can communicate when their concepts and ontologies do not align beforehand.
Research & advanced systems
Research that becomes working systems.
My research work is centered on intelligent agents, semantic communication, negotiation, and grounded AI workflows. I use implementation as part of the research process so protocols, assumptions, and failure modes can be inspected rather than remaining only conceptual.
Research themes
Protocol-driven agent interaction
Making negotiation and coordination inspectable through explicit states, message flows, and constrained dialogue.
Grounded AI systems
Designing workflows where generated conclusions remain traceable to source evidence and intermediate processing stages.
Observable agent architecture
Treating tracing, state transitions, and communication paths as core engineering concerns rather than debugging afterthoughts.
Selected research systems
Three projects I can explain from problem to architecture.
Flagship · Multi-Agent Systems
STRATON-LLM
An ontology-based communication and negotiation system for heterogeneous intelligent agents. STRATON-LLM explores how agents can detect semantic mismatches, negotiate shared meaning, and persist accepted mappings without relying on a pre-built common ontology.
Multi-Agent SystemsOntologiesRabbitMQConsulFastAPILLMs
View case study →Open Multi-Agent Framework
LATTICE Framework
A domain-agnostic framework for decentralized collaboration between specialized autonomous agents, with explicit routing, communication, state ownership, memory, knowledge, and observability contracts.
PythonAsyncIOMulti-Agent SystemsMessage BrokerLLM Agents
View case study →Grounded AI · Research Tooling
Research-Paper Deconstruction Agent
A source-grounded workflow for extracting, normalizing, structuring, and analyzing technical papers so implementation guidance remains tied to the supplied research rather than unsupported model knowledge.
PythonLangGraphLLMsPDF ProcessingGrounded AI
View case study →How I approach research engineering
I prefer explicit system boundaries, inspectable state transitions, reproducible experiments, and source-grounded evaluation. The objective is not to make a research prototype look complicated; it is to understand exactly where the system succeeds, where it fails, and why.