Mahvash Siavashpour
About Me
I am a research assistant at Amii and the University of Alberta, working with Dr. Michael Bowling. My research interests include conditional compute in neural networks, continual learning and cognitive science.
I have 5+ years of experience spanning academic Machine Learning research, applied industry projects especializing on Large Language Models, NLP, Reinforcement Learning, and agentic AI.
Education
- M.Sc. in Computer Science, University of Alberta, 2023-2026
- B.Sc. in Computer Engineering, Amirkabir University (Tehran Polytechnic), 2018-2023
Experiences
Associate Machine Learning Developer at AltaML.
Built agentic LangChain systems for document processing, created a real-time LLM-as-a-judge evaluation pipeline, and shipped a production RAG chatbot.
Graduate Research Assistant at University of Alberta.
Benchmarked a surgical triage task with LLM decision-makers, improving ROC-AUC via fine-tuning, DPO post-training, chain-of-thought reasoning, and synthetic data generation.
Work Integrated Learning Opportunity (WILO) at Alberta Machine Intelligence Institute (Amii).
Co-authored technical reports, translated ML concepts for non-technical partners, and proposed AI solutions to startups across healthtech, vision, and forecasting.
Undergraduate Research Assistant at Amirkabir University of Technology.
Designed a transformer-based sequence labeling model for joint punctuation insertion and sentence-boundary detection with baseline benchmarking.
Research Intern at Institute for Cognitive and Brain Sciences.
Modeled human lexical decision-making using machine learning, evidence-accumulation models, and Bayesian inference.
Publications
-
A Multimodal Deep Learning Ensemble Framework for Building a Spine Surgery Triage System, 20th World Congress on Medical and Health Informatics
-
Semi-Automated Construction of Sense-Annotated Datasets for Practically Any Language, 31st International Conference on Computational Linguistics (COLING 2025)
-
Developing a triage model for surgical referral from primary care based on spinal radiology reports using natural language processing (NLP) and large language models (LLMs), The 19th International Back and Neck Pain Research Forum 2025
-
UAlberta at SemEval-2024 Task 1: A Potpourri of Methods for Quantifying Multilingual Semantic Textual Relatedness and Similarity, In Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024)
-
Modeling Human Lexical Decision-Making with Artificial Neural Networks and Evidence Accumulation Models, Proceedings of the Annual Meeting of the Cognitive Science Society 2023
Licenses & Certifications
DLRL 2025
Issuer: Amii (Alberta Machine Intelligence Institute)
Issued: Aug 2025
Credential ID: 157687298
Credential: Certificate
Work Integrated Learning Badge
Issuer: Amii (Alberta Machine Intelligence Institute)
Issued: Nov 2024
Credential ID: 124653940
Credential: Certificate
Powered by Jekyll and Minimal Light theme.