kevin@castro: ~/portfolio — zsh
Energy

Rowland.ai

Benchmarked extraction pipeline across 10,000+ lease scenarios

$ open project →

Lead developer on an AI-powered document extraction platform for the energy sector, owning the server-side systems, data pipelines, cloud infrastructure, and API layer the platform runs on. I architected the LLM pipelines, RAG infrastructure, and tool-calling systems behind autonomous, multi-step document reasoning using Gemini’s long-context models, and built the Golden Corpus validation framework: automated recursive testing with divergence alerting that continuously checks extraction output against verified benchmarks.

I stress-tested the bulk-processing pipeline across 10,000+ lease scenarios under enterprise document loads, and mentored engineers on backend system design and AI tooling while acting as the technical liaison between product and engineering on roadmap decisions.

$ cd .. back to work
technologies
  • Terraform
  • Node.js
  • TypeScript
  • React
  • Next.js
  • PostgreSQL
  • Supabase
  • S3
  • Azure
  • Google Cloud
  • PostHog
  • Claude
  • Gemini
  • Vertex AI
  • FastAPI
skills
  • Document Processing
  • DevOps
  • CI/CD
  • Agentic LLMs
  • OCR
  • API Development
  • System Design
  • Infrastructure as Code
  • Database Design
  • JavaScript
  • Python