Nathan Rumsey

Engineered corrections to the short-rate cancellations in billing engine, implementing transactional logic using Gosu to prevent erroneous refunds and safeguard $100k+ in annual revenue.

Refactored a legacy VB.NET actuarial regulatory reporting tool into C#, introducing parallel bulk processing and unit test coverage to reduce turnaround time from days to hours.

Unblocked new customer acquisition by resolving an address validation issue in the issuance workflow that prevented foreign customers from purchasing policies, enabling seamless onboarding and avoiding manual workarounds.

Developed BDD scenarios with Cucumber/Gherkin to validate core business workflows, eliminating one week of manual validation and ensuring on-time regulatory compliance for high-stakes releases.

Extended C# .NET test infrastructure for a precision motion control system, leveraging dependency injection to validate against both simulated and physical hardware and improving test result reporting.

Designed a field failure tracking system and data consolidation pipeline for laser components, processing 10k+ records spanning decades to deliver reliability insights and RCA to leadership, informing warranty projections.

Engineered Python CLI tool to compress and archive AI computer vision defect detection data from the manufacturing floor, reducing storage needs by 90% and saving $3000+ annually in hardware costs.

Automated quality metrics reporting across cross-functional teams using Azure dataflows and Power BI backed by Azure Data Lake, eliminating manual entry errors and saving leadership 10+ hours monthly.

Engineered a stateful agentic system that orchestrates multi-step reasoning loops through a custom MCP server, managing persistent character state and ensuring deterministic tool execution for rule adjudication.

Designed a hybrid retrieval engine combining structured SQL state with vector-based RAG over a comprehensive rulebase, enabling the agent to perform dynamic constraint satisfaction and validate action legality.

Led a 5-developer Agile team building a scalable React/Express application on AWS, directing sprints and technical architecture while implementing accessibility via UX prototypes and automated testing for WCAG compliance.

Architected a distributed backend deployed to containerized AWS infrastructure, supporting 10k+ concurrent users under load testing; built CI/CD pipelines to strengthen release reliability.

Engineered a hybrid search engine combining keyword and vector-based semantic retrieval using custom-trained embeddings and FAISS, achieving 0.06s latency and 10x faster results than Windows Search for 100k+ files.

Designed a modular, event-driven AI indexing pipeline with a Redis worker queue, SQLite metadata store, and incremental updates enabling near-real-time synchronization of file changes via semantic embeddings.

Built a multi-source energy analytics platform on Databricks using a full medallion architecture, standardizing energy, weather, and solar data into curated Delta tables with conformed dimensions and hourly/daily fact models.

Exported curated analytical tables to a dedicated S3 publishing bucket and integrated them with Snowflake using a minimal-IAM storage integration, enabling a Streamlit dashboard to query all data through Snowflake.