Portfolio
Work
Ship fast, stay legible, keep humans in the loop. Architecture and execution — with a psychology-informed standard for what the product actually does to the humans using it.
You^n.tech
Principal AI Architect
I'm best when…
The AI works in a demo, but fails under real-world usage — trust, reliability, clarity.
The timeline is aggressive and the team needs architectural ownership, not decks.
The product needs to be explainable and human-centered without losing technical rigor.
Current focus
Principal AI Architect · Multi-agent OS · 2021–Present
You^n — identity-aware, explainable multi-agent operating system
Problem
Most AI tools bolt intelligence onto existing workflows without rethinking the system. The result: agents that can't explain themselves, recommendations nobody trusts, humans stuck correcting what they can't understand. The hard problem isn't building an AI — it's building one a real person can actually rely on.
Constraints
- End-to-end ownership: architecture → data model → UX → deployment
- Explainability and auditability required as first-class architectural constraints
- Human-in-the-loop by design — not as an afterthought
Results
- Production AI OS built around human legibility as a first-class constraint
- Full audit trail on every agent decision — no black boxes
- End-to-end architecture and execution from zero to production
What I did
- Designed multi-agent orchestration with explicit responsibility boundaries per agent
- Built identity-aware reasoning layer for context and workflow alignment
- Implemented transparent scoring and decision audit trails throughout
- Owned the full stack: orchestration, data models, frontend UX, backend, deployment
Founder · Multi-tenant Agent Platform · 2024–Present
NanoClaw + ProductOps — production multi-agent platform with 14-skill ecosystem
Problem
AI agents are easy to demo and hard to operate. Running them reliably at scale — across multiple clients, channels, and workflows — requires a runtime layer that most teams don't build until they're already in trouble. The goal was a platform that could operate like a senior engineer at 3am: autonomous, accountable, and recoverable.
Constraints
- Multi-tenant isolation: each client runs in their own containerized context
- WhatsApp-first UX — agents must be accessible from conversational interfaces, not dashboards
- 14 skills must compose cleanly into end-to-end workflows without context collision
Results
- End-to-end agentic platform: trigger from WhatsApp → full site built and deployed
- Zero-to-live site workflow runs in a single orchestrated session
- 14-skill ecosystem composable into arbitrarily complex multi-step product workflows
What I did
- Architected NanoClaw — WhatsApp-first multi-tenant agent runtime (Docker, Claude Agent SDK)
- Built ProductOps — 14-skill system: /icp, /build-site, /deploy, /qa, /seo, /image-gen, and more
- Designed skill sequencing, memory isolation, and stateful workflow orchestration
- Implemented multi-channel delivery: WhatsApp, Telegram, Slack, and web interfaces
Head of AI · Voice + RAG · HumanOp · 2024–2025
HAi — voice RAG agent MVP delivered in <8 weeks
Problem
Deliver a voice-based retrieval-augmented agent that answers accurately and interacts in a way that feels coherent, grounded, and usable — not just technically correct.
Constraints
- 0→1 delivery under an aggressive timeline
- RAG systems require precision and failure-mode awareness from the start
- User trust and clarity are part of the technical definition of done
Results
- MVP delivered from zero to production in under 8 weeks
- Polymorphic RAG routing live — structured, unstructured, and voice-context paths
- Cross-team enablement through structured internal AI bootcamps
What I did
- Delivered the HAi voice RAG MVP in <8 weeks
- Developed the Polymorphic RAG Vector-based Human Authentic Intelligence engine
- Built guardrails and trust-oriented interaction patterns into the core loop
- Ran AI bootcamps for internal teams: Operations, R&D, and Marketing/Sales
Track record
Research · Simulation · Evaluation · 2025
Coherence Benchmarking Lab — simulation benchmarks for quantum coherence topologies
Problem
Agentic AI systems need evaluation frameworks that go beyond accuracy. The Unity Pixel Framework explores a deeper question: how do information topologies — the shape of how data and state propagate through a system — affect coherence, reliability, and emergent behavior at scale?
Constraints
- Novel evaluation methodology — no existing benchmark to build from
- Results must be reproducible and interpretable, not just numerically interesting
- Simulation runs expensive: design must minimize compute while maximizing signal
Results
- Tile topology outperformed linear chain at high step counts — coherence advantage documented
- Reproducible benchmark run published with full JSON artifact
- Unity Pixel Framework theory documented and available as a public 1-pager
What I did
- Designed the Unity Pixel Framework — a tiling-based coherence model for information propagation
- Built simulation benchmarks comparing tile vs. linear chain topologies at scale
- Ran comparative benchmarks across 240 steps with controlled seed for reproducibility
- Published theory overview and benchmark data as open artifacts
Artifacts
Simulation Benchmarks of Quantum Coherence
Tile vs. linear chain topologies — peer-reviewed benchmark paper
The Unity Pixel Framework
Theory overview — information topology and coherence design principles
Benchmark Run — Tile vs Linear, Seed 5
240-step simulation artifact (JSON) — full reproducible run data
Founder & CEO/CTO · WebAR SaaS · 2017–Present
AR Hero — no-code WebAR platform with measurable conversion lift
Problem
Build an in-house WebAR SaaS that turns product visualization into actionable behavior — making 'cool AR' into measurable conversion impact for brands without developer resources.
Constraints
- Lean stack and pragmatic implementation choices — bootstrapped
- Ship features that move conversion metrics, not novelty
- SaaS realities: iteration speed, reliability, and customer adoption
Results
- Bootstrapped SaaS shipped and in production
- Programmable AR CTAs lifted client conversion measurably
- Lean Firebase architecture still running — 7+ years in production
What I did
- Built a no-code WebAR SaaS from scratch on lean Firebase stack
- Implemented programmable AR call-to-action overlay layer for brands
- Designed for ease of use: clients A/B test AR CTAs without touching code
Director of Engineering · Platform Consolidation · 2015
Evolving Wisdom — CMS consolidation supporting $6M → $16M growth
Problem
Unify fragmented content systems so the organization could operate faster, publish reliably, and scale revenue without platform drag creating a bottleneck to growth.
Constraints
- Four legacy CMS platforms to consolidate without disrupting operations
- Keep the business running and stakeholders aligned during major change
- Coordinate delivery across a distributed, global team
Results
- 4 → 1 CMS consolidation completed
- Revenue growth from $6M → $16M supported by platform stability
- 12+ concurrent programs delivered with a global team
What I did
- Consolidated 4 legacy CMSs into a single WordPress stack
- Managed 9-person global team across 12+ concurrent programs
- Streamlined publishing and commerce workflows to reduce operational friction
Director of Technology · Commerce · Sounds True · 2008–2015
Sounds True — 10,000+ SKU migration and subscription MVP in 10 weeks
Problem
Modernize delivery and monetization for a leading conscious media publisher: migrate a massive audio catalog to cloud and ship subscription capability under an aggressive deadline.
Constraints
- 10,000+ SKUs with high correctness requirements
- Aggressive 10-week delivery window for subscription MVP
- Must set a durable, maintainable foundation for recurring revenue
Results
- 10,000+ SKUs migrated to cloud delivery
- Subscription MVP shipped on time under aggressive deadline
- Development budget paid back within 6 months of launch
What I did
- Migrated 10,000+ audio SKUs from CD to cloud delivery infrastructure
- Shipped subscription mobile app MVP in 10 weeks
- Prioritized correctness, reliability, and long-term maintainability throughout