Arezoo Alipanah

Arezoo Alipanah

ML Researcher @ UWaterloo

Reinforcement Learning & AI Safety

LLMs & Knowledge Graphs

About Me

Researcher, Programmer, Thinker, Creative Doer

Research Associate @ University of Waterloo (Cheriton School of CS)

Machine learning researcher (MASc, University of Waterloo) specializing in Reinforcement Learning, AI Safety, and Large Language Models. I am currently working with Prof. Pascal Poupart on combining knowledge graphs with LLMs for alignment-oriented applications in AI for Education.

My MASc thesis focused on Safety and Robustness in Inverse Reinforcement Learning, where I developed algorithms to identify and exploit vulnerabilities in IRL agents — resulting in a first-author publication at the Canadian Conference on AI (2025).

I am passionate about theory-driven research with practical implementation. More specifically, I work on:

1. Inverse Reinforcement Learning & AI Safety

2. Knowledge Graphs + LLMs for reducing hallucinations (RAG)

3. AI Agent Safety & Orchestration

Motto (some of my favourite book quotes)

"What is a person, if not the marks they leave behind?"

V.E. Schwab

The Invisible Life of Addie LaRue

"So we'll train, Feyre, until the last possible day. Because we never know if just one extra hour will make thedifference."

Sarah J. Maas

A Court of Wings and Ruin

"We've all got both light and dark inside us. What matters is the part we choose to act on. That's who we really are."

J.K. Rowling

Harry Potter and the Order of the Phoenix

"The people who are crazy enough to think they can change the world are the ones who do."

Steve Jobs

"Think different" Commercial

"Somewhere inside all of us is the power to change the world."

Roald Dahl

Matilda

Specialities

ML Researcher

Focused on Reinforcement Learning, AI Safety, and LLMs. I thrive on theory-driven research with practical implementation — from IRL robustness to knowledge graph pipelines for education.

Programmer

Expert Python developer with hands-on experience in PyTorch, TensorFlow, HuggingFace, and building RL pipelines from scratch. Coding is my comfort zone.

Educator

Teaching Assistant at UWaterloo (LLMs, Transformers, Algorithms). Former Stanford Code in Place Section Leader. Passionate about making complex topics accessible.

Reader & Writer

Always eager to dive into new worlds — people usually find me with a book in hand or an audiobook in my ears. I also write about my research, books, and ideas on this blog.

AI Ethics Advocate

As an Associate at Amii's AI Career Accelerator Program, I help design and facilitate hands-on workshops on AI Ethics and responsible AI development.

Multilingual

Fluent in Persian, Azeri, and English. Being multilingual has shaped how I think about communication, knowledge transfer, and building bridges across cultures.

Affiliated With:

University of Waterloo
Amii
Code in Place, Stanford
Shahid Beheshti University
K.N. Toosi University

Certifications & Badges

Amii AI Career Accelerator – Work Integrated Learning

AI Career Accelerator
Work Integrated Learning — Amii

Amii AI Ethics 1: Governance

AI Ethics 1: Governance
Amii & Digital Governance Council

Amii AI Ethics 2: Deployment

AI Ethics 2: Deployment
Amii & Digital Governance Council

Stanford Code in Place Section Leader

Section Leader
Code in Place — Stanford University

Fun Facts

Publications

4

Students Mentored

100+

Books Read

300+

Years of Study

10

Resume

Education

2023–2025

M.A.Sc. in Electrical & Computer Engineering (Pattern Analysis & Machine Intelligence)

University of Waterloo, Waterloo, ON

Cumulative GPA: 92.2 / 100.0

Thesis: Safety and Robustness in Inverse Reinforcement Learning

2019–2022

M.Sc. in Mechanical Engineering, Dynamics & Control

K. N. Toosi University of Technology, Tehran, Iran

GPA: 17.87/20

2015–2019

B.Sc. in Mechanical Engineering

Shahid Beheshti University, Tehran, Iran

Experience

Sep 2025 – Present

Research Associate

Cheriton School of Computer Science, University of Waterloo

Advisor: Prof. Pascal Poupart

Knowledge Graphs + LLMs for AI in Education; AI agent safety & orchestration; RAG pipelines to reduce hallucinations.

Sep 2023 – Sep 2025

Graduate Research Assistant

CL2-Group, University of Waterloo

Advisor: Prof. Yash Vardhan Pant

Safety & robustness in Inverse RL; adversarial demonstration attacks on IRL agents; first-author publication at Canadian AI 2025.

Jan 2024 – Sep 2025

Teaching Assistant

University of Waterloo, Dept. of ECE

ECE 657 Machine Intelligence: lectures on Transformers & LLMs.

ECE 406|606 Algorithm Design: algorithm analysis and complexity.

Feb 2020 – Sep 2022

Graduate Research Assistant

ARAS Lab, K. N. Toosi University of Technology

Deep RL for robot path planning in dynamic environments (reward shaping); published at ICRoM 2022.

Apr 2021 – May 2024

Section Leader (Volunteer)

Code in Place, Stanford University

Facilitated weekly Python discussion sections for students worldwide (CS106A equivalent).

AI & Deep Learning

PyTorch

Reinforcement Learning / Inverse RL

LLMs & Transformers (HuggingFace)

Knowledge Graphs

TensorFlow

OpenAI Gym

Languages & Tools

Python (Expert)

MATLAB

C++

SQL

LaTeX

Git

Download CV

Portfolio

Some of My Works

Blog

My Thoughts
ARLAS: Adversarial RL for Agent Safety
Mar2025

ARLAS: Teaching AI to Fight Itself Safe — A Visual Explainer
BrowserGym & AgentDojo Workshop
Apr2025

BrowserGym & AgentDojo: Hands-On Workshop on Web Agents & AI Safety
what happened to you
2Aug

What Happened to You; A Book for Everyone
Being A Girl in Academia
20Jun

Being A Girl in Academia- An Ali Hazelwood POV

Contact

Get in Touch

Get in Touch

I am open to research collaborations in Reinforcement Learning, AI Safety, LLMs, and Knowledge Graphs. I'm also happy to chat about mentoring, tutoring (Python, RL theory and code), or exciting AI projects.

Feel free to reach out — I try to respond within a few days. The contact form below goes directly to my inbox.

Waterloo, ON, Canada
aalipana@uwaterloo.ca

Contact Form