Jethro Arciaga

I’m an AI/LLM Engineer and Python Developer with hands-on production experience building LLM-powered automation on top of Claude, including prompt evaluation, structured extraction, and classification pipelines that process real-world, unstructured data at scale.

My background in data engineering (ETL, PDF mining, web scraping, AWS) helps me build AI systems with solid data-pipeline foundations—not notebook-only prototypes. I’m comfortable across the full loop: prompt design, evaluation/benchmarking (Promptfoo), MCP server integration, and shipping to production with CI/CD and automated testing.

At GovConnex, I architect and scale AI-driven data extraction and ingestion systems with LLM automation as a core focus. I built an end-to-end asynchronous pipeline using Claude to ingest incoming mail, extract payloads, and dynamically classify message types in production, and I integrated Promptfoo to benchmark and optimize prompt performance while eliminating regressions before release. I also designed advanced PDF mining and built both a low-latency “Live-Transcript” streaming pipeline and an “Instant-Transcript” engine driven by Python and web scraping.

Before that, I built and maintained ETL pipelines with client integrations on Azure Blob Storage, using Selenium and BeautifulSoup, plus secure API workflows. Earlier, I worked as a backend developer on a client web application, writing AWS Lambda functions, managing APIs and permissions with AWS Cognito, operating a Neo4j graph database, and scheduling report exports via AWS EventBridge.