Research
Published work on multi-agent AI systems: identity-aware protocols, structured collective reasoning, and the infrastructure that makes delegation verifiable.
Papers
Trust infrastructure for multi-agent AI.
Identity, delegation, provenance, and collective reasoning, with experiment code and open-source reference implementations. Honest null results included.
Agent Identity Protocol: Verifiable Delegation Across MCP and A2A
In a scan of ~2,000 public MCP servers, none enforced authentication. OAuth 2.1 covers single-hop auth; AIP adds Invocation-Bound Capability Tokens for multi-hop delegation chains with cryptographic scope attenuation.
0.049ms verification 100% adversarial reject 0.086% chain overhead Project hub Paper Blog post CodeAn Identity-Aware Protocol for Multi-Agent LLM Systems
Rich delegate identity cards, progressive payloads, governed sessions, and trust domains enable metadata-aware routing: easy tasks to fast models, hard tasks to capable ones.
~12x lower latency 37% fewer tokens 96% attack detectionFrom Debate to Deliberation: Structured Collective Reasoning
Typed reasoning moves, disagreements preserved as first-class objects, and a convergence algorithm that guarantees termination in bounded rounds.
+0.95 vs debate 9.56 hidden-profile 62x cost ceilingThe Provenance Paradox in Multi-Agent LLM Routing
When delegates inflate self-reported quality, quality-based routing selects the worst delegates, worse than random. Delegation contracts and attested identity fix it.
9.51 attested routing 36 configs confirm sub-µs overheadOpen Source
Protocol implementations and experiment harnesses.
aip
Agent Identity Protocol. Verifiable, delegable identity for AI agents across MCP and A2A. Rust + Python reference implementations. Apache 2.0.
ldp-protocol
LLM Delegate Protocol — Python SDK and Rust reference implementation. Identity-aware routing, provenance tracking, trust domains. pip install ldp-protocol
ldp-research
Experiment code and data for the LDP papers. Six research questions, A2A baselines, ablation conditions, LLM-as-judge evaluation.
enterprise-rag-bench
RAG patterns benchmarked for enterprise. Five chunking strategies, five retrieval patterns, evaluation harness, guardrails.
applied-nlp-research
Production NLP from the pre-LLM to post-LLM era. Capsule networks, BiLSTM-CRF for NER, transformer fine-tuning, PyTorch.
Profiles
Academic profiles.