Hello! I'm
I ship LangGraph + Bedrock agents that replace ClickOps and RAG systems that serve 5,000+ docs over Microsoft Teams. I govern it all with HCP Terraform and Sentinel policy-as-code across 200+ AWS accounts and 50+ Azure subscriptions — so the AI has safe, auditable infrastructure to run on.
Open to roles in Dallas, TX · or remote
About Me
Senior AI Cloud Platform Engineer with 9 years across cloud, infra, and automation. At Royal Caribbean I turned a 3-day ClickOps process into a 10-minute self-service agent — and built the production RAG platform and IaC governance layer it runs on. I work where agentic AI meets enterprise infrastructure.
Tech I work with
AWS Solutions Architect
AWS AI Practitioner
Azure Administrator
AWS GenAI Developer
Terraform Associate
Projects
Enterprise RAG Chatbot
Production RAG reference architecture on Amazon Bedrock Knowledge Bases with Aurora pgvector and Titan v2 embeddings. Claude Haiku handles query reformulation and sub-query decomposition, hybrid semantic + keyword retrieval surfaces candidates, Cohere Rerank v3.5 reranks via cross-encoder, and Claude Sonnet produces grounded answers with citations. Ships with a Comprehend-based PII sanitizer and DynamoDB audit trail.
query → Haiku rewrite → hybrid search → Cohere rerank → Sonnet + citations
Agentic AWS Account-Vending Platform
LangGraph + Bedrock (Claude) agent in Microsoft Teams that validates intent against live AWS Organizations state, auto-assigns CIDRs, and triggers HCP Terraform runs with human-in-the-loop approval. Cut account provisioning from 3+ days to under 10 minutes for 10+ teams.
Autonomous AI Agent Fleet
9 autonomous AI agents on AWS Lambda using Strands SDK with Bedrock (Claude Sonnet/Haiku) and Gemini fallback. Shared common library handles model-tier routing, DynamoDB dedup, Discord delivery, and CloudWatch observability. 100% Terraform-managed, Python uv-workspaces monorepo.
Internal MCP Server for LLM Tool-Calling
Model Context Protocol server exposing AWS Organizations, Terraform state, and HCP run APIs as typed tools. Standardizes auditable agent workflows across the vending platform and future AI use cases.
Parallel Terraform Import Tool
Python tool (asyncio, uv workspaces) on GitHub Actions matrix pipelines that discovered 45k+ live resources across 200+ AWS accounts and auto-generated import blocks plus scaffolded HCL, migrating the org from ClickOps into version-controlled IaC.
Writing
Building a Production RAG Platform on Bedrock Knowledge Bases
Aurora pgvector, Titan v2 embeddings, Cohere Rerank v3.5, and a CI-integrated eval harness with LLM-as-judge scoring. What worked, what didn't, and the retrieval metrics that actually moved.
Coming soonAgentic Account-Vending with LangGraph + HCP Terraform
How a 5-tool LangGraph agent collapsed AWS account provisioning from 3 days of ClickOps to 10 minutes of self-service in Microsoft Teams — with human-in-the-loop approval and policy-as-code guardrails.
Coming soonWhy You Should Build an Internal MCP Server
Wrapping AWS Organizations, Terraform state, and HCP run APIs as typed MCP tools turns every LLM agent into a first-class, auditable platform citizen. Architecture, auth model, and lessons learned.
Coming soonGet In Touch
I'm actively looking for AI Cloud Platform, AI Infrastructure, or Senior Platform Engineering roles — open to positions in Dallas, TX or fully remote. If you're hiring or just want to connect, my inbox is always open.