Amir Tavasoli

Toronto, Canada

Amir Tavasoli

Director of Data Science at Sobeys Inc., leading AI strategy across loyalty and marketing for one of Canada's largest grocery retailers. I build the decision layers, operating models, and leadership systems that turn machine learning into something executives can trust and scale.

View Case Studies Download Resume Contact
Portrait of Amir Tavasoli

Executive Focus

Applied AI leadership, agentic AI build and scaling, process change through AI agents, personalization strategy, experimentation design, and delivery systems that can stand up to executive scrutiny.

About

Read Full Bio

After studying Computer Science and completing a Master's focused on Natural Language Processing, I began my career building forecasting and optimization models at Canadian Tire. I later applied and expanded that experience in the eCommerce environment at Home Depot. At Paytm Labs, I moved into leading machine learning engineering teams, gaining deeper exposure to the technical and organizational requirements for scaling AI/ML systems in production.

These experiences positioned me to transition from individual contributor to leading teams of data scientists and ML engineers at Sobeys. There, I led large-scale personalization systems serving millions of customers while managing cross-functional delivery across engineering, marketing, and analytics. That progression ultimately led me into a director role, expanding my scope into enterprise marketing AI and advancing initiatives in agentic AI to modernize decision-making across the organization.

My work centers on the layer between models and business execution - designing operating models, prioritization frameworks, and measurement loops that turn analytical outputs into repeatable, scalable business outcomes.

Experience

Sobeys Inc. — Director of Data Science

Feb 2024 – Present

  • Lead all Advanced Analytics initiatives within the Marketing organization, expanding scope from Data Science Manager to enterprise-wide AI leadership.
  • Build and scale AI, ML, and Agentic AI solutions that directly support marketing strategy, including:
    • One-to-one personalization
    • Generative AI use cases
    • Retail media measurement
    • Marketing Mix Modeling (MMM)
    • Churn prediction
    • Personalized pricing
    • Customer life stage management
    • eCommerce activation
  • Design and architect end-to-end AI systems in partnership with cross-functional teams, ensuring alignment between technical architecture and business objectives.
  • Own the full lifecycle of solutions, from data ingestion and modeling to deployment, experimentation, and executive decision support, ensuring scalability, reliability, and measurable impact.
  • Lead development of two generative AI agents:
    • An automated, context-aware marketing offer generation and personalization agent
    • An AI decision layer over MMM outputs to optimize channel allocation and marketing investment
  • Implement a production-grade MMM framework integrating multi-agency data pipelines and linear mixed models to generate elasticity curves and guide spend optimization.
  • Guide Retail Media Network measurement strategy by:
    • Building robust A/B testing pipelines
    • Developing Markov-based multi-touch attribution models
    • Quantifying cross-channel customer interaction impact across owned and third-party platforms
  • Enforce rigorous, state-of-the-art measurement methodologies across programs to ensure transparency, accountability, and repeatable business outcomes.

Sobeys Inc. — Data Science Manager, Personalization

Jul 2021 – Feb 2024

Promoted to Director

  • Led a 15-person full-stack ML team, across hybrid and offshore staffing, delivering enterprise personalization across Sobeys banners.
  • Oversaw weekly recommendations for 16M customers across email (SFMC), web, and mobile platforms.
  • Owned the end-to-end pipeline, from Enterprise Data Lake extraction through model development, deployment, and measurement.
  • Architected and reviewed scalable ML systems using PySpark and Databricks in Azure while enforcing code quality and production standards.
  • Implemented MLOps and reliability practices to improve automation and stability.
  • Partnered with Marketing, Loyalty, and Cloud teams, managed agile delivery through JIRA and PlanView, and provided executive reporting on ROI, costs, and program health.
  • Delivered 75% cloud cost reduction, 95% fewer pipeline failures, 4x scale growth, 10x incremental sales lift, and a Bayesian framework for dynamic campaign measurement.

Paytm Labs — Technical Lead & Senior ML Engineer

Jul 2020 – Jul 2021

  • Built and maintained scalable ML systems for personalized recommendations, in-app advertising, and marketing optimization across hundreds of millions of users.
  • Developed production-grade models using Scala and Python with AWS-based infrastructure.
  • Designed and maintained data engineering pipelines to clean and transform large-scale application log data into ML-ready features.
  • Created executive dashboards and visualizations supported by Kubeflow-based ML workflows.
  • Operated in a dual role as both Machine Learning Engineer and team lead within the core ML team.
  • Managed and mentored a team of 6 MLEs, translating business requirements into technical roadmaps and delivery plans.
  • Led agile planning through JIRA, coordinated cross-functional stakeholders, and produced executive-level progress and impact reports.

Home Depot — Data Science in eCommerce

  • Developed deep learning-based recommendation engines using TensorFlow and PyTorch within Google Cloud Platform.
  • Deployed recommendation systems across site personalization, digital advertising, and email customization.
  • Enhanced search relevance through NLP-based ranking and personalization algorithms.
  • Conducted research to evaluate and implement state-of-the-art modeling approaches.
  • Collaborated cross-functionally to automate workflows and improve operational processes.
  • Worked closely with the central Data Science team in Atlanta to align modeling standards and deployment practices.

Canadian Tire — Lead Data Scientist in Retail Modeling

  • Implemented large-scale fuzzy text matching for millions of automotive part numbers in a Hadoop environment using MapReduce (Java) with Spark backend.
  • Developed and deployed machine learning solutions across multiple retail use cases, including:
    • Automotive parts forecasting (including cold-start items)
    • Seasonal product forecasting (e.g., pools)
    • Related item identification
    • Lifecycle analysis
    • Market basket analysis
  • Built models primarily in IBM SPSS and Python using Random Forest, XGBoost, and time-series methods (SARIMA, Holt-Winters).
  • Provided technical leadership for enterprise store-network initiatives such as sales forecasting, lost-sales estimation, space optimization, and price optimization.
  • Designed end-to-end technical architectures, including object-oriented Python frameworks, PySpark project structures (SparkML), and distributed ML optimization tools using DASK and Docker.
  • Managed Agile delivery processes (SCRUM in JIRA), coordinated cross-functional stakeholders, and reported progress to VP and SVP-level leadership.

Education

2003-2008

B.Sc. Computer Science

Amirkabir University of Technology (Tehran Polytechnic), Tehran, IR

Average: 17.13 / 20.00, 2nd highest grade in the program.

2008-2010

M.Sc. Computer Science

McMaster University, Hamilton, CA

Thesis topic: Automated Message Triage: A Proposal for Supervised Semantic Classification of Messages. This work used a combination of text mining algorithms on messages exchanged between patients and their physicians, implemented in R and Java using LingPipe and Apache Server.

Supervisor: Dr. Norm Archer

Publications & Presentations

2026

Tavasoli, A., From Models to Decisions: Agentic AI for Marketing Mix Optimization, Talk at the Atlantic AI Summit 2026, University of New Brunswick, Fredericton, NB.

June 2005

Qualified for the World Finals in the RoboCup 2005 Robotics Competition, Rescue Simulation League, Osaka, Japan.

2020

Tavasoli, A., Journey of a Data Scientist: Using AI, Machine Learning and Big Data to Solve Problems in the Retail Industry, Talk at Lakehead University, Thunder Bay, Ontario.

2013

Tavasoli, A., Archer, N. P.: Automatic message triage: A decision support system for patient-provider messages, Americas Conference on Information Systems 2013, McMaster University: Hamilton, ON. This paper uses text mining to triage text messages exchanged between patients and their physician.

2009

Tavasoli, A., Archer, N. P.: A Proposed Intelligent Policy-Based Interface for a Mobile eHealth Environment, Innovation in an Open World. G. Babin, P. Kropf and M. Weiss, Springer Berlin Heidelberg. 26: 246-251. This paper presented at MCETECH 2009, and it is focused on an adaptable interface for novice users in mobile devices.

2009

Presentation of eHealth Integration System using HL7 v3, IBM CASCON Technology Showcase November 2009. Proposed a service-oriented architecture to translate XMLs generated using two health standards.

Let's Connect

I am open to conversations about executive roles, advisory work, speaking, and building data science organizations that produce measurable outcomes. If there is a mandate to make applied AI more reliable, more accountable, or more commercially effective, I am interested.