Steven Rice

Tech Stack

Projects and Publications

A Novel Hybridized Forecasting Technique Utilizing ARIMA and Large Language Models

• Proposes a multi-agent forecasting system including two large language model agents.
• Demonstrated up to an 18.84% improvement in success rate and total failures were reduced by up to 38.28% over an ARIMA-only model in a COVID-19 case study.

Solving Inverse Kinematics using Large Language Models

• The first known work to propose utilizing large language models to solve inverse kinematics problems.
• The proposed methodology successfully solved kinematic chains of up to three degrees-of-freedom.
• Conducting a follow-up work leveraging chain-of-thought methodologies and incremental solving.

Fusion IK

Solving Inverse Kinematics using a Hybridized Deep Learning and Evolutionary Approach

• The first deep learning method for inverse kinematics that did not require an existing dataset, as this methodology builds its own to train the neural network.
• The methodology showed improvements in solving and movement time on extended kinematic chains.

Easy-AI

• Teaches students sensors and actuators, finite state machines, steering behaviors, pathfinding with corner-graph node placement and A* navigation, and multi-agent behaviours.
• Used to teach "COMP-4770 Artificial Intelligence for Games" at the University of Windsor.

Vector Search Engine

• Developed in .NET C# using Blazor with a Qdrant database and embeddings generated with word2vec.
• Provides concise summaries generated with Llama 3.2 and similar document searching.

Dynamic Duos

• A capture-the-flag game with weapons, pickups, and map rotation designed to be a reference for students.
• The template provides server hosting and joining, a lobby and ready-up system, and character movement.
• Used to teach "CSSE-4900 Game Design, Development, and Tools" at the University of Detroit Mercy.

Asteroids Reinforcement Learning

• Demonstrates how to use reinforcement and imitation learning to create intelligent agents.
• Used to teach "COMP-4770 Artificial Intelligence for Games" at the University of Windsor.

Virtual Monitors

• Prototype virtual and augmented reality system to display desktop monitors in Unity.
• Can create and render multiple virtual monitors for Windows.

Facial Expression Recognition

• Solution for the "Kaggle Challenges in Representation Learning: Facial Expression Recognition Challenge".
• Trains ResNet18 with a gradual addition of data augmentation.

Staff Management Software

• Allows for easy shift logging by staff and approval by managers.
• Developed in .NET C# using Blazor, ASP.NET Identity, Entity Framework, and a SQLite database.
• The Town of Tecumseh used it for seasonal staff.

Networking Autograder

• Tested, graded, and provided feedback for student submissions that implemented a streaming protocol.
• Developed in Python with FastAPI and a Supabase database.
• Used to teach "COMP-3670 Computer Networks" at the University of Windsor.

Experience

Adjunct Professor

• Instructor for "CSSE-4900 Game Design, Development, and Tools", a course focusing on Unity.
• Taught students engine architecture, C# scripting, processing input, character movement, physics and collision handling, user interfaces, networking, and artificial intelligence.
• Designed the course content, assignments, and examinations.

Graduate and Teaching Assistant

• Created an artificial intelligence library for Unity to teach "COMP-4770 Artificial Intelligence for Games".
• Created the final project and an automatic grading server for "COMP-3670 Computer Networks".
• Taught artificial intelligence concepts, game development, networking, algorithms, and data structures.

Software Developer

• Created a robotics and vision simulation platform with Unity to develop digital twins.
• Advised the company during meetings with Unity Technologies.
• Developed C# applications in ASP.NET MVC and WPF.
• Created responsive user interfaces for applications by integrating DataTables.

Information Technology Specialist

• Automated the company's order confirmation process with Python, OpenCV, Tesseract OCR, and Selenium, reducing task completion time from hours of manual work to minutes.
• Led development on the company's first ASP.NET project.
• Developed and optimized databases, backends, and frontends.
• Developed lightweight translation capabilities for websites.

Pool Coordinator

• Responsible for the training, scheduling, and supervision of a staff of twenty members.
• Developed software to automate shift logging and wage calculations, greatly improving efficiency.
• Managed daily facility operations.

Lifeguard and Swim Instructor

• Ran lifesaving, leadership, swimming, and Aquafit classes.
• Responded to first aid situations.

Conferences

Windsor-Essex Google Developers Group DevFest 2024

Build With AI

• DevFests are Google-sponsored software developer conferences that allow the local software development community to network and learn about the coolest new technology.
• Presented how my research at the University of Windsor is proposing novel applications of large language models in areas they are known to be weaker in such as robotics and forecasting.

52^nd North American Manufacturing Research Conference (NAMRC 52)

• SME's North American Manufacturing Research Conference is recognized as North America's preeminent and longest-running international forum for applied research and industrial applications in manufacturing and design.
• Presented my research "Fusion IK: Solving Inverse Kinematics using a Hybridized Deep Learning and Evolutionary Approach".
• Discussed digital twinning and simulations with other researchers, offering them guidance on their Unity-based projects.

Education

Master of Science in Computer Science

Artificial Intelligence Stream

Honours Bachelor of Computer Science

Artificial Intelligence Specialization

Honours Bachelor of Commerce Business Administration

Minor in Mathematics

Awards