Abdallah Zubedi — Data scientist & data engineer, Toronto, Canada

This study follows a single subject across ten years, three continents, and two disciplines. The subject began in marketing, developed an unusual attachment to measurement, and now builds the systems that do the measuring: scheduled intelligence agents, a collection pipeline that reads 10,000+ websites, and a sync, still in build, to keep a contact database honest. Observed effects include a +600% lift in clicks per send, a +40% lift in qualified lead volume, and a persistent habit of photographing wildlife. The instruments turn out to be indifferent to the industry: a pipeline that reads manufacturer websites would read sensor logs the same way. The pattern holds. Replication is invited.

Findings

  • 4–6 hrs manual research saved per week Weekly intelligence is a pipeline, not a reading list. Four research agents, one per weekday beat (competitors, industry bodies, market signals), run on a schedule in GitHub Actions. Each reads its memory of prior runs, crawls the week, passes every claim through a fact-check gate that demands a source URL, and renders the briefing as an HTML email to the team. Replaced four to six hours of manual research a week.
  • 10,000+ client websites crawled for files Ten thousand websites is a loop, not a workload. Automation that visits the websites of 10,000+ building-product manufacturers, locates the BIM, CAD and specification files on each, and uploads them into the company's platform, where they used to arrive one at a time by hand.
  • 10 min sync interval, CRM to email platform Two databases agree only if something makes them. A polling pipeline, in build, to keep the contact database and the email platform in agreement. Every ten minutes it reads what changed, routes each contact to the right lists by tag, moves prospects to clients when the tag changes, and never re-subscribes anyone who opted out. The full base is backfilled once, in bulk; the pipeline carries the deltas from there.
  • +600% clicks per send Email engagement responds violently to systematic testing. Redesigned and systematically tested the HTML email templates (subject lines, layout, send timing) across 114 campaigns. Clicks went from about 100 to about 700 per send.
  • +40% qualified leads Lead quality is a data-quality problem, not a sales complaint. Automated the lead-logging and reporting workflows so a booked meeting became a qualified-lead record in real time, with validation at the point of entry (about 30% less time spent checking data). Cleaner, faster data surfaced better prospects and lifted qualified-lead volume roughly 40%.
  • 2nd place, national Strategy survives contact with judges when it is built on data. A food-waste-reduction strategy for the Loblaw Innovation Challenge: market research, financial modelling, and consumer data behind a multi-scenario plan. Placed 2nd nationally, among Canadian Master's cohorts.

Experience

  • Kuala Lumpur, Malaysia (2016 to 2020) Treatment period one. BSc Marketing at Lancaster / Sunway (2:1), a peer-reviewed article, "Assessing Consumer Skepticism Towards Sustainable Marketing" (Journal of Marketing Management and Consumer Behavior, 2020), its fieldwork set in the energy sector, and a first data role at Tapway: 140+ qualified leads pulled from overlooked segments.
  • Nairobi, Kenya (2020 to 2021) Treatment period two. Marketing and business development at Eccentric: +18% regional FMCG sales through gamified activations, and a data-driven omnichannel funnel that lifted conversions ~30%.
  • Toronto, Canada (2021 to present) Treatment period three. Master of Management (MMIE) at Queen's, GPA 4.11/4.3, 2nd place in the Loblaw Innovation Challenge. Then Communitech, Moonstone, and AEC Daily, where the subject stopped only measuring systems and started building them.

Skills

SQL (daily use), Python (pandas, scikit-learn, statsmodels), Data pipelines (ETL, validation, scheduled jobs), APIs & web scraping (REST clients, crawlers at 10,000+ sites), LLMs & AI agents (agent playbooks, fact-check gates), Git & CI automation (GitHub Actions, scheduled workflows), R (ggplot2, tidyverse), Visualization (dashboards, hand-rolled SVG), A/B testing (design and analysis), Attribution (regression-backed), Marketing (B2B, where it started)

Education

  • BSc (Hons) Marketing, Lancaster University, Malaysia campus (2016–2020)
  • Master of Management (MMIE), Queen's University, Kingston, Canada (2021–2022)
  • MSc in Analytics (Computational Data / Data Science), Georgia Institute of Technology, Atlanta, USA (2026–present)

Publication

Zubedi, A. T. A., Nezakati, H., & Abdi, L. (2020). "Assessing Consumer Skepticism Towards Sustainable Marketing." Journal of Marketing Management and Consumer Behavior. Archived copy.

Contact

dalla_zubedi@outlook.com · linkedin.com/in/abdallahzubedi