I build systems that let software reason, act, coordinate, verify, and settle.
My work sits across:
- agentic and LLM systems
- programmable data and intelligent infrastructure
- cryptographic authority and execution control
- verification and evidence systems
- onchain coordination and settlement
I tend to work below the application layer, turning recurring problems into reusable primitives, protocols, and infrastructure.
Programmable clearing and settlement infrastructure for work, commerce, and agentic services.
A commercial request becomes a bounded, verifiable payment lifecycle: Workspace → delegated authority → agreement (Pact) → proof → USDC settlement → verifiable Receipt.
Context-aware execution control for autonomous data changes.
iGraph turns live organizational context into bounded execution authority, enforcing signed Impact Pacts before an autonomous system can make consequential changes — and emitting machine-readable Change Receipts after.
An experimental oracle architecture for turning LLM-derived claims into independently evaluated, evidence-backed onchain attestations.
Inference → independent node evaluation → commit/reveal → quorum → settlement.
Agentic opportunity intelligence built around verified capability, provenance, evidence gaps, and approval-gated execution.
Norn ranks grants, bounties, and partnerships against a verified capability profile — distinguishing evidence from claims, and never acting without explicit approval.
A protocol specification and reference implementation for atomically coordinating multiple financial intents through signed Pacts, deterministic commitments, execution controls, settlement, and recovery semantics.
I am interested in the primitives underneath intelligent systems:
Authority — who is allowed to act, under what mandate, within what boundary?
Evidence — what happened, what proves it, and how can another system verify it?
Execution — how does an autonomous system move from intent to a bounded real-world action?
Coordination — how can independent actors or agents work together without giving up control?
Settlement — how does the system reach a final, auditable state?
The goal is not to add AI to existing software. It is to build the infrastructure that makes autonomous software trustworthy enough to actually operate.
Agents · LLMs · Programmable Data · Protocols · Verification · Cryptographic Authority · Settlement · Distributed Systems