Ali Parsaee

Ali Parsaee

Machine Learning Researcher

I am a research who loves to work on new and challenging problems in the world of Machine learning, Math, or Health AI. I often enjoy challenges that come from learning in settings where data is limited or very noisy, investigativng causal relations using ML techniques, or in optimizing gathering new data via active learning or bayesian optimisation methods. I also love researching LLM agents and swarms.

About Me

I am a Machine Learning Researcher with expertise in developing innovative AI solutions. I obtained my bachelors in Biological sciences at the University of Calgary in 2021 and my masters in Computing Sciences at the University of Alberta in 2025 specializing in Statistical Machine Learning. My thesis work regards finding near optimal way to gather and use data in hard settings (with limnited information) using Active Learning, Optimisation, and Survival analysis.

More recently I have been working in areas of transfer learning, as well as using LLM agents in efficient ways. I enjoy collaborating with diverse teams and continuously learning new techniques and frameworks.

Work Experience

Nov 2024 - Present

Machine Learning Resident

Alberta Machine Intelligence Institute (AMII)

Researched Bayesian optimization and transfer learning techniques for life sciences applications. Designed models to improve media formulation for cell growth across multiple cell types.

Apr 2024 - Sept 2024

Data Science Consultant (WILO)

Alberta Machine Intelligence Institute (AMII)

Developed scalable ML pipelines using SQL and Python. Delivered prototype ML applications and actionable insights to industry partners.

Apr 2024 - Sept 2024

Educator on working with Genetic data

Alberta Machine Intelligence Institute (AMII)

Work Integrated Learning Oppportunity (WILO) where I researched and helped compose a Coursera course teaching ML scientist effective dimension reduction techinques for using Genentic datasets.

Apr 2023 - Nov 2024

Graduate Research Assistant

University of Alberta

Built statistical models and time series analyses to predict disease diagnosis timelines. Developed optimal algorithm for data gathering in survival settings under budget constraints.

Apr 2023 - Dec 2023

Machine Learning Intern

University of Alberta

Engineered data pipelines and developed classification models for alloy properties prediction, achieving 90% accuracy.

Sept 2021 - Aug 2024

Python and Data Analysis Instructor

Calgary Chess School

Designed and taught Python and data visualization curriculum. Created open-source educational materials for students and the broader community.

Research Work

Synchronous vs Asynchronous RL in a Real World Robot

Presented at RISE Conference

Developed and compared synchronous vs asynchronous updates in soft actor-critic algorithms using the FRANKA robotic arm.

Read Paper

Alzheimer's and its Relation to Herbal Medicine

Published in *Biology*, MDPI

Conducted a meta-analysis of Alzheimer's mechanisms and the role of herbal compounds in altering disease pathways.

Read Paper

Wisdom of the Machines: Exploring Emergent Intelligence in LLM Crowds

Submitted to ACM

Investigates the emergence of intelligence through the aggregation of LLM agents like Qwen2 and LLaMA 3.2 using a new mathematical framework.

Read Paper

Crohn's Disease Survival Prediction

Submitted to Bioinformatics

Predicts time to first bowel resection in pediatric Crohn's patients using clinical and lab data. In collaboration with medical professionals.

View Report

Budget-constrained Active Learning to Effectively De-censor Survival Data

Submitted to ICML

Proposes a novel data acquisition algorithm guaranteeing a 63% optimal convergence rate under budget constraints. Validated on medical and financial datasets.

Read Paper

LoopBench: Discovering Emergent Symmetry Breaking Strategies with LLM Swarms

Submitted to ANTS conference

Proposes a novel benchmark in evaluating how llm agents collaborate and deal with complexity via np hard problems.

Read Paper

Projects

BioHazardGPT

AI-powered safety classification tool for chemistry and biology labs. Offers real-time risk assessment (Safe, Caution, Hazardous) with contextual reasoning using both rule-based and GPT-4o RAG systems.

Technologies: GPT-4o, RAG, FAISS, ChromaDB, Streamlit, GitHub Pages, Python

Live Demo View Code

Pocket AI Therapist

Runner-up at NatHacks Hackathon. Built a multi-modal AI model for real-time sentiment and emotion analysis using speech and facial cues.

Technologies: OpenCV, Hugging Face, Amazon Polly, Python

View Project

Country Clustering

Used K-means clustering on Happiness Report data to visualize socio-economic trends among countries.

Technologies: Streamlit, Pandas, Scikit-learn, Python

View Project

Contact Me