Founder · operator · applied-AI builder · MBA, Pepperdine
Founder of ZONA. I enjoy building consumer products and architecting the revenue systems underneath them. Most recently, I've shipped applied AI products into real operations, in the US and abroad, to help businesses grow.
🌐 unknownbehavior.com · 💼 LinkedIn
- Founder of ZONA, a real-time nightlife platform that turns user behavior and demand into where tonight is. I lead product, design, and engineering across iOS and Android, plus the venue portal, ads marketplace, promoter portal, ticketing, and lifecycle engine. 150K+ downloads, 25+ venue partners, live since June 2024.
- Building with AI across the stack: a lifecycle decision engine with A/B-tested creative, plus standalone systems like Sentra (runtime control for AI agents), Cortex (dual-model review), and voice-coach (a real-time voice agent).
Product & growth
- ZONA. Consumer nightlife marketplace plus B2B partner platform (portal, ads, promoter, ticketing). Product, brand, monetization, and partner delivery end to end. More at unknownbehavior.com.
Applied AI & systems
- sentra. Runtime execution control for autonomous AI agents: scores risk, enforces allow/block/shutdown. Model-agnostic.
- sentra-medication. Runtime governance for AI in medication: validates AI-generated orders against live patient data (FHIR R4) and returns allow/block/escalate in under 100ms, built to EU AI Act Article 12. An NYU Berlin team project I led; we presented to Google and Join Capital and won the final pitch before a faculty panel. Live demo.
- cortex. Dual-model governance: a worker writes, an overseer stress-tests, agents respawn with memory after repeated failures.
- voice-coach. A real-time AI voice agent you talk to out loud: streaming speech-to-text → Claude → text-to-speech in one low-latency pipeline (Pipecat, Deepgram, Cartesia, Silero VAD), with live tool-use and a swappable personality.
- care-gap-engine. Population-health outreach prioritization: ranks open care gaps by clinical urgency × response likelihood × equity, then drafts personalized patient outreach with Claude (prompt-cached), behind a public methodology card. Live demo.
- autonomous-claims-workflow. Multi-agent workflow on IBM watsonx.ai + Granite, governed end to end by Sentra (IBM SkillsBuild AI Lab, 2026).
- symptom-triage-coach-v2. Multi-modal LLM build: image + text in, schema-validated output, with an eval harness.
Earlier data & analysis
SQL and Python over public datasets: medicare-billing-analysis, ed-wait-time-dashboard, healthcare-access-risk-analysis, clinical-trial-trends.
- Product & growth: product strategy, partner platforms, marketplace systems, lifecycle/CRM, monetization & pricing, A/B testing
- Applied AI: multi-agent systems, real-time voice agents (Pipecat · Deepgram · Cartesia · Silero VAD), multi-modal & vision (Claude), tool-use / function-calling, LLM evals & eval harnesses, prompt caching, LoRA/QLoRA fine-tuning, Hugging Face, PyTorch, Claude API, IBM watsonx.ai
- Engineering: Python, Node.js, Express, FastAPI, Streamlit, MongoDB, Firebase, React Native, REST APIs
- Data: SQL, Python (pandas, NumPy), experimentation, dashboards
- Languages: English / Spanish