AI software engineer
designing agentic systems
that hold up in production.
I design, build, and operate AI systems used by real teams — intent routing, safe tool-use, structured outputs, and the serverless infrastructure underneath. I treat AI as production software, not a demo.
Production AI and the systems around it.
A short list — production AI agents and the personal builds that sharpen the same instincts. The full list lives on the projects page.
A Microsoft Teams bot powered by AWS Bedrock and n8n that lets enterprise users trigger backend operations through natural conversation — Lambda routing, structured JSON decisions, confirm-before-invoke on destructive actions.
Natural-language query agent for PeopleSoft on Bedrock + Lambda, with effective-dated row handling. Gives non-technical users direct access to HR and payroll data without writing SQL.
Turns plain-English prompts into working apps. Auth, billing, sandboxed code execution, real-time preview.
Interactive visualizations for sorting, searching, and pathfinding — built to make algorithms feel obvious instead of abstract.
Self-hosted Linux server running Dockerized Node and Flask services behind Nginx, with SSL and SSH hardening. Live in production.
Where my attention is.
Building
Production-grade AI agents — a Microsoft Teams assistant on AWS Bedrock that gives users a safe, conversational interface to backend systems, and a Text-to-SQL agent over an enterprise HR / payroll database. Plus migrating a decades-old Smalltalk/FileMaker stack onto serverless AWS.
On personal time: LingoGrasp (cross-platform language-learning app) and an AI SaaS generator — auth, billing, sandboxed code execution, real-time preview.
Going deep on
The patterns that make agents reliable in production: multi-step tool use, structured outputs, schema-grounded prompting, and confirm-before-invoke on destructive actions. Field notes land in the Learning Lab.
Closing out an M.S. in Software Engineering (AI focus) and the AWS Solutions Architect / Developer associate track.