Professional Experience

Research, full-stack engineering, analytics, and production systems.

What's This?

My professional experience connects software engineering with data and AI. I have built LLM agent workflows for physics research, museum web applications, analytics pipelines for retail forecasting, and backend services for logistics operations. For questions or collaboration ideas, email me at kevin.you@mail.utoronto.ca.


Research Assistant @ University of Toronto AI Physics & Safety Lab

  • Developed LLM-based optimization pipelines using Gemini-1.5-Flash and GPT-4o APIs to automate advanced physics simulations, achieving a 20% relative performance gain on real-world physics problem solving.
  • Designed AI agent systems with TextGrad, Automated Design of Agentic Systems, AutoGen, and LangChain to optimize compound AI workflows and improve zero-shot GPQA accuracy to 67%.
  • Integrated a RAG workflow with RAGflow and Ollama so agents could access an internal physics literature knowledge base, reducing factual hallucinations in simulation design by 20%.

Information Technology Intern @ Alpha Education

  • Led development of the museum website with React.js and Node.js, implementing interactive Home and Exhibitions features that increased website views by 30%.
  • Engineered two interactive map-based applications with Leaflet.js and Mapbox API, incorporating 100+ historical events and engaging 20% of visitors after deployment.
  • Optimized website performance with lazy loading and code minification, reducing average page load time from 2.1 seconds to 1.3 seconds.

Data Scientist Intern @ Nestlé

  • Developed and optimized Spark MLlib regression and classification pipelines for demand forecasting, promotional spend optimization, and revenue growth, reaching 72.6% R-squared accuracy.
  • Engineered automated ETL pipelines in PySpark on Databricks, transforming 5,000+ raw records from heterogeneous sources into structured Delta tables for downstream analytics.
  • Implemented a feature importance framework with Gradient Boosting Regressor to identify sales performance drivers and support promotional strategy decisions.

Software Development Engineering Intern @ ThirstyBrain.Inc

  • Developed a logistics management system with Java and Spring Boot, providing RESTful APIs to track and optimize delivery routes across 1,500+ daily shipments.
  • Improved request throughput by 45% by integrating Redis route caching and Kafka event streaming between microservices.
  • Deployed containerized applications to AWS using Docker and Kubernetes, enabling auto-scaling and high availability under variable traffic.
  • Achieved 90% JUnit test coverage across services and database layers while following Agile practices for iterative delivery.