Most AI systems fail at execution. I build the control planes that make them accountable.
Control-Plane Architect for AI Systems
Building real-time decision systems, agentic infrastructure, and correctness frameworks.
I work on systems where AI must act, not just predict — trading, data integrity, and on-device intelligence.
-
AI Decision Infrastructure
Bounded-latency decision loops, evaluation pipelines, and observable execution systems -
Agentic Control Planes
Systems that coordinate tools, workflows, and outcomes with measurable behavior -
Correctness & Data Integrity (ASA/DCML)
Architectures for verifiable storage, arbitration, and recovery under failure -
On-device AI Systems
Swift/SwiftUI + CoreML pipelines for privacy-preserving, real-time inference
Real-time decision system using multi-signal consensus and regime-aware execution
→ https://phoenix.industriallystrong.com
Dual-chain mirrored ledger for verifiable storage and recovery
→ https://industriallystrong.com/lab
Labs, experiments, and system proofs across AI, storage, and physical computation
→ https://industriallystrong.com
A control plane for long-running work with LLMs, built on evidence-bound claims, layered authority and explicit conflict resolution
→ claw-public
- Reczipes2: recipe photo → Vision OCR → LLM extraction → structured recipes, with allergen, FODMAP and GutSense analysis
- qrl-architecture-comparison: side-by-side demo of an LLM-only pipeline against a decision architecture (MHT and FAISS regime retrieval)
- KanjiKanaTrainer: handwriting practice for kana and Chinese characters, with PencilKit stroke capture and stroke-order scoring
- KeepTrack: an iOS daily intake tracker with goals and Siri capture
- litho-shared-contracts: shared Swift, TypeScript and JSON Schema contracts for a computational lithography decision engine
- https://industriallystrong.com
(Control planes, AI systems, correctness, and physical computation)