Oliver Thompson

Engineered a self-healing multi-agent system using LangGraph to automate renewable grid anomaly classification, improving incident response times by 48%. Deployed an enterprise RAG pipeline across 40,000 regulatory compliance documents using vector databases, reducing research delays to under 5 seconds. Built real-time scraping and parsing workflows for highly volatile energy spot prices, achieving 99.95% data accuracy. Led the migration of manual diagnostic report writing into a secure, LLM-generated output framework, saving 80+ staff hours monthly. Focused on reliable orchestration, retrieval quality, and backend automation performance.

Programmed and launched an autonomous solar asset tracker that continuously parsed telemetry feeds to predict hardware maintenance cycles. Designed custom API connectors and implemented automated API integration to link ML forecast datasets directly into corporate ERP software, increasing scheduling efficiency by 22%. Fine-tuned open-source LLMs (Llama-2/Mistral) for domain-specific engineering text processing, reducing reliance on external LLM API integration costs by 35%. Converted stakeholder efficiency goals into modular Python scripts delivered through bi-weekly agile sprints, and built supporting FastAPI services and integrations.

Developed scheduled backend automation scripts and server-side scripting to clean and ingest national hydrological flow data, saving 18 hours of manual labor weekly. Built internal business process automation tooling using Zapier and custom webhooks to remove onboarding bottlenecks. Optimized legacy server-side automation code to reduce database script execution overhead by 30% across local networks. Co-engineered automated data validation checks that blocked malformed IoT device logs, supporting continuous compliance of data pipelines.