geoffrey robinson

I’m a Software Engineering Intern focused on building production-grade systems that stay reliable under real load. At Husky Voice AI, I architected an appointment management system handling 1000+ appointments daily and implemented call log CSV exports using Redis queues and background batch processing.

I also contribute to open-source data orchestration and RAG frameworks. In CocoIndex & Cognee, I migrated to a type-safe StableKey system to reduce serialization/runtime errors, shifted incremental processing tracking tables into custom Postgres schemas, and decreased relational pipeline data processing latency by 200% using a schema-only ingestion pipeline with httpx for high-throughput asynchronous fetching.

My project work keeps me close to end-to-end product outcomes—from inference speed to scalable persistence. I developed a production-ready Email Phishing Detection MCP Server that enables LLMs with real-time phishing classification via a custom API, optimized inference speed by exporting a Logistic Regression model to ONNX for sub-10ms predictions in Node.js, and deployed the Dockerized service to Render for high availability.

I’m currently pursuing a Bachelor of Engineering in Information Science & Engineering (expected September 2026), and I bring a “ship and improve” mindset to every build. I’ve also built InkGrader with pgvector-powered similarity search for millisecond-latency rubric matching and CiteGeist using Django REST Framework and React for interest-based research recommendations—plus I earned a 1st Place win (Contribute To Win 2025 Fall 2025) through my Cognee contributions.