I've built TruEval, a Python LLM evaluation harness for repeatable quality control across OpenAI, Anthropic, Gemini, and local models. Its rule-based and LLM-as-Judge evaluators detect regression after prompt or model changes, with SQLite baselines and HTML reporting.
I also built SlimPrompt, a RAG prompt-compression middleware that cuts token usage up to 80% using LLMLingua-2, semantic deduplication, and auto-calibrated compression ratios. My work spans RAG pipelines, vector databases, QLoRA fine-tuning, and agentic workflows.
At Taylored Analytics, I design and maintain multi-cloud data pipeline integrations across GCP, Cloud Storage, and AWS. I build Python/Flask/PostgreSQL and React integrations for AI-ready enterprise data pipelines and financial-services reporting workflows.
I'm double Google Certified as a Professional Cloud Architect and Professional Data Engineer, with hands-on delivery across GCP, AWS, Docker, Kubernetes, CI/CD, and full-stack applications.
