I architect platforms that handle real load — multi-tenant SaaS, identity systems, microservices at scale. Now bringing that discipline to LLM ops and agentic AI.
I've spent 25 years architecting platforms that had to handle real load — multi-tenant SaaS, identity and SSO at scale, event-driven microservices, plain old enterprise Java that couldn't go down at 2am.
Now I'm bringing that discipline to AI. The hard part of agentic systems and LLM ops isn't the model — it's the boring infrastructure: context, retries, observability, evals, deploy. The same problems I've been solving for decades, with a new failure mode on top.
I help teams move from "AI demo" to "AI in production" without breaking what already works.
Identity & SSO powering learning products serving millions of students. Token issuance, federation, multi-tenant authorization. Spring Boot · Kafka · Redis · Kubernetes.
Multi-tenant CRM with omni-channel messaging. ~1,500 commits, NestJS microservices on Kubernetes, Kafka backbone, multi-region. Sub-second SLAs on inbox routing.
Operational SaaS for queues, scheduling and field operations. Architecture from blank repo — backend, infra, identity, deploy pipeline, billing. Sold and still running.
First public work in LLM ops, agentic patterns and production AI infra. Open-source experiments and write-ups landing through 2026.
Brown belt in Brazilian jiu-jitsu, skydiver, father of three. The same constancy that ships systems.
