Maria Chzhen

University of Toronto.

prof_pic.jpg

Hello! I am an undergraduate Industrial Engineering student at the University of Toronto, specializing in AI/ML.

My core interest lies in applying robust AI/ML, operations research, and optimization techniques to create intelligent solutions for real-world problems like energy systems, sustainability, and geospatial applications.

Currently, I’m developing novel machine learning approaches for handling strict, complex constraints in power systems as a research student with Prof. Priya Donti (MIT EECS & LIDS). Alongside my research, I am also an AI intern at Rocket Innovation Studio (Architecture team), contributing to the development of AI-driven solutions for Rocket’s business needs. At UofT, I have worked with Prof. Florian Shkurti (Vector Institute, UofT Robotics Institute) on perception algorithms implementation.

I’m driven by the challenge of designing intelligent systems that are not only effective but also safe and reliable, especially within critical domains.

education

uoft

University of Toronto

BASc. in Engineering | 2023-2028 (expected)
Industrial Engineering Major, Artificial Intelligence Minor
Activities: Academic Outreach Director at IEEE UofT, VP Student Academics at SEA UofT.

academic research

Power systems project Research in Safe and Constrained Reinforcement Learning
[Research Project] @ Donti Lab (MIT LIDS).
In progress. Developing new machine learning techniques to address challenges of safety and complex constraints in sequential decision-making, with potential applications in energy systems.
Mechanistic Interpretability Interpreting Internal Representations of Transformer Models
[Research Project] @ UofT Math Directed Reading Program. With Daniel S., Kiarash S., Vincent C., and Charles S.
In progress. Investigating the use of sparse autoencoders to interpret internal representations and features learned by large language models.
Implicit Occupancy and Flow Fields
[Research Project] @ RVL (UofT).
Implementation of the algorithm proposed in the Implicit Occupancy and Flow Fields for Perception and Prediction in Self-Driving paper for mobile robotics.
PDSI Comparison of Regression Algorithms in Multivariable Drought Modeling and Analysis
[Research Project] Bronze medal at Canada Wide Science Fair 2023. Oral presentation at the 23rd AMS Conference on Artificial Intelligence for Environmental Science.
[Link to project]
Used and compared classical machine learning algorithms to study the relationship between drought and various climate factors in diverse regions around the world.
NDVI Multiple Factor Regression Analysis of Vegetation Changes and Land Degradation in Drylands
[Research Project] Third place at the International Science and Engineering Fair 2022, Earth and Environmental Sciences category. Poster presentation at the 23rd AMS Student Conference.
[Link to project]
Used multiple regression to understand the relationship between different environmental factors (like climate, human activity, and soil properties) and changes in vegetation cover and land degradation in dryland regions.

technical projects

Resume screening Fair and Interpretable Resume Screening System
[Internship Project] @ Rocket Innovation Studio.
[Demo coming soon!]
End-to-end system for resume screening. Optimization-based approach (inspired by optimal sparse decision trees) to build a candidate screening engine providing transparent and verifiable decision logic. Full-stack application built with React and Node.js.
Light pollution map Light Pollution Map
[Internship Project] @ Rocket Innovation Studio.
[Demo coming soon!]
AI-powered interactive map to help cities reduce light pollution and improve the quality of life for residents. Integration of an AI assistant to provide actionable insights and recommendations for reducing light pollution. Exportable reports for city officials to make informed decisions.
Ocean plastic collection Reinforcement Learning for Strategic Ocean Plastic Retrieval
[Project] @ Earth Menders. Presented at the 2025 Canadian Undergraduate Mathematics Conference (CUMC).
[Presentation link]
Reinforcement Learning (Proximal Policy Optimization) to optimize the strategic collection of ocean plastic by autonomous agents.
Edge AI project Edge AI for Real-Time Audio Recognition: A Voice-Controlled Sound System
[Club Project] @ IEEE UofT. With Adam O., Katherine Y., Grace S., and Athavan B.
[Demo coming soon!]
Co-developed an edge AI tool for real-time sound recognition. The device recognizes speech and plays different sounds based on the user's command. Implemented and deployed a real-time Whisper model on a Raspberry Pi with embeddings to match the user's command to the correct sound.
Gesture detection project Gesture detection app
[Club Project] @ IEEE UofT. With Athavan B. and Scott H.
[Demo coming soon!]
Co-developed a gesture detection tool that plays different sounds based on the user's gesture. Implemented a real-time model to match the user's gesture to the correct sound.