Solution Architect
Solution architect with a builder mindset. 15 years of Java, Spring Boot, and AWS: enough depth to know which patterns survive contact with production. Now designing AI-native enterprise systems end-to-end: voice-first claims pipelines, document intelligence at scale, agentic orchestration with Temporal and LangGraph. Coding agents close the loop - the velocity of a team, the judgment of a veteran.
Full-stack developer on an OLAP analytics platform for AC Nielsen. Java backend, front-end UI, data analysis support. Where I learned that the boring infrastructure work is actually the important work.
Enterprise Java across regulated industries: State Street (regulatory reporting, REST services, in-memory cache for large datasets), NY Life (ACCORD XML transformation via SOAP/Apache CXF), MetLife (batch-to-realtime migration).
Started building core policy and agency management systems in Java/Spring Boot on Kubernetes. Now architecting AI-native platforms: voice-first claims reimagination (Temporal + LangGraph), document intelligence pipelines (AWS + Gemini), and production agentic systems with Bedrock and MCP tooling.
Designing systems where deterministic and AI steps each own their domain - architecture sanity over AI theater. Making the hard calls: what Temporal owns, what LangGraph owns, where the audit trail must live. The judgment that stops AI sprawl before it becomes technical debt.
15 years in insurance and finance means compliance and audit trails are designed in - not bolted on after an incident. Every AI step in a regulated system needs an escape hatch, a paper trail, and a rollback path. The differentiator most AI architects lack.
Senior architects used to have a gap between what they could design and what they could build alone. Coding agents close it. I architect the system and drive implementation velocity that used to require a full team - with the senior judgment to catch what the agents get wrong.
PoC to production is where most AI projects go quiet. 15 years of integration work - REST, SOAP, Kafka, GraphQL, legacy systems - means I know what the handoff to operations looks like. AI systems that can't survive enterprise integration review don't ship.
The hardest decision in Claims Reimagination wasn't which LLM to use. It was where the AI boundary sits. Temporal owns lifecycle durability. LangGraph owns cognitive reasoning. When those domains are clear, the whole system is debuggable.
15 years in insurance and finance means compliance, audit trails, and failure modes are designed in, not bolted on after an incident. Every AI step in a claims system needs an escape hatch and a paper trail.
The prod support agent started with workflow tools and human-in-the-loop steps. Agents that do too much too fast are hard to debug and harder to get sign-off on. Trust is a product feature.
Most production AI problems are data quality, latency, and integration complexity - not model capability. I've spent more time on output validators, retry logic, and API rate limits than on prompt engineering.
Senior architects used to have a gap between what they could design and what they could build alone. Coding agents close it. I can architect the Temporal + LangGraph orchestration layer and drive implementation velocity that used to require a team. The judgment still has to be yours. That's what keeps the generated code from accumulating hidden debt.
Open to conversations about AI-native system design, enterprise agentic architecture, or the intersection of senior engineering depth and AI velocity. Not actively looking but happy to talk.