Ali Mahdipour-Shirayeh

Ali Mahdipour-Shirayeh, PhD

I am a Principal Data Scientist at Roche/Genentech-Data Science Acceleration Team. Leveraging Gen AI, LLM, and diverse ML techniques like Causal ML, Reinforcement Learning, and gradient boosting, Spark, cloud computing along with CI/CD pipelines, I tackle business challenges, providing actionable insights for product efficacy, process automation and optimization for clinical studies and health authority platforms.

Before attending Roche, I worked at Precisely-AI Innovation Lab in which my job, as a Senior Data Scientist, was to use machine learning, deep learning, LLMs and cloud services to develop anomaly detection pipelines, automate data quality systems, and AI-readiness platforms for structured/unstructured data sets. My role was tied with optimizing decision-making, enhancing efficiency, and driving real-time insights.

Prior to that, I was a Data Scientist II at Manulife-Advanced Analytics & AI where I developed and deployed advanced machine learning and AI solutions, including Gen AI, LLMs, RAG, causal ML, and gradient boosting for fraud detection, risk assessment, customer segmentation, virtual health coaching, and recommender systems. I have also implemented a distributed training framework, formulated A/B test plans for ML models, and led teams of data scientist and data engineers to create fraud detection models. I have also developed a real-time analytics dashboard, automated ML processes, and collaborated across teams to enhance efficiency and scalability.

Previously I was a data scientist at University of Toronto and Princess Margaret Cancer Centre, University Health Network. I utilized Machine Learning/AI approaches, Statistical Analysis and Mathematical Modeling to develope/design end-to-end pipelines to study cancer patients' big data. I aimed to capture crucial features and provide an optimal representation of data and then developing appropriate data science/machine learning systems to study, validate, continusouly monitoring and optimizing the system.

We have used data science and machine learning to identify clonal ancestry and minimal residual disease for clinical patient relapse which suggested novel classifications and better prediction of the disease. The results of our projects provided a step forward towards a better understanding of biological background of diverse cancer which likely provide a step forward towards more intense therapeutic strategies and ultimately control/cure for cancer. Previously, I was a Postdoctoral Fellow and then Data Scientist and Scientific Associate I in this lab. My resaerch interests are Data science, Statistics, Machine Learning, NLP, Deep Learning and Gen AI/LLM.

I completed my PhD program in Biomedical Data Science at the Applied Mathematics Department of the University of Waterloo under the supervision of Prof. S. Sivaloganathan and Prof. M. Kohandel . My thesis title was "Evolutionary Dynamics of Cancer: Spatial and Heterogeneous Effects" .

As a researcher, I am always eager to learn and pursue, examine and develop cutting-edge discoveries as well as to increase collaboration opportunities and productive scientific worldwide connections. I actively contribute to AI research, staying updated with emerging trends and best practices.

Selected Publications

Multiple Myeloma B Cells and Pre-Plasma Cells Are Important Reservoirs for Myeloma Relapse Following Plasma Cell-Directed Therapy and Prevent Cure with Standard Therapies

Blood 142 2023

Single-cell profiling of multiple myeloma reveals molecular response to FGFR3 inhibitor despite clinical progression

Cold Spring Harb Mol. Case Stud. 2023

sciCNV: High-throughput paired profiling of transcriptomes and DNA copy number variations at single cell resolution

Briefings in Bioinformatics 2022

Gain(1q) promotes mitochondrial oxidative phosphorylation and suppresses interferon response and tumor immunity in multiple myeloma and other human cancers

Clinical Lymphoma Myeloma and Leukemia 2021

Modeling cell dynamics in colon and intestinal crypts: The significance of central stem cells in tumorigenesis

Bulletin of Mathematical Biology 2018

Genotype by random environment interactions gives an advantage to non-adaptive neutral minor alleles

Nature Scientific Reports 2017