Jul 2026 — Present
Founding Data Scientist
Book Depot Inc. · Thorold, Ontario
I own pricing intelligence across a 60K+ SKU catalog: elasticity, causal DML, inventory pacing, and the unglamorous work of making those models survive a real catalog.
About
Who I am
I am Hassan Mehmood, an AI/ML Engineer & Quant Systems Architect. I live in the part of the work where a model has to become a decision: a price, a match, a flag, a posture. I like that part more than the demo.
I have about three years of shipping ML, data platforms, and BI in fintech and B2B shops. Before that I taught. I still think like someone who has to explain the thing out loud.
My journey
Health sciences first. Then markets, analytics, models, and the systems that keep models from becoming folklore. I did not plan the path as a brand. It is just what kept being interesting.
- Business Analytics ↓
- Data Science ↓
- Machine Learning ↓
- ML Engineering ↓
- AI Systems ↓
- Quantitative Finance ↓
- Building
What I build
Sep 2025 — Jul 2026
Data Scientist
Iteration Matrix · Remote
Production ML for supplier mastering, ROI attribution, and audit recovery. Reporting errors went from 25% to under 5%. I also owned the MLOps path: CI/CD, containers, retraining, drift, monitoring.
May 2025 — Aug 2025
Data Science Analyst
EQ Bank · Toronto, Ontario
Automated ETL and live Power BI for model performance. Cut manual reporting by about 90%. Built a multithreaded Polars quality system over more than 1TB of SQL Server data.
Sep 2022 — Aug 2024
Instructional Specialist, Data Analytics & FinTech
University of Toronto · Remote
Mentored 50+ students through Python, SQL, machine learning, and time series. I like teaching the parts people usually skip: why the model is lying, and what to do next.
What I'm learning
Tighter identification in pricing. Evaluation for agents that is not theatre. How much of a quantitative system can stay local and still be serious.
What I'm interested in
Causal ML, production systems, capital markets as a measurement problem, and teaching people who are done collecting certificates.
Currently thinking about
Causal ML that has to price real things
Elasticity is easy to estimate and hard to trust. I keep coming back to confounding, inventory, and the gap between a notebook and a catalog.
Agents as systems, not demos
A loop that calls tools is not an architecture. I am more interested in memory, evaluation, and what happens when the agent is wrong on a Tuesday.
Markets as a measurement problem
Macro regimes, analogues, and position sizing. Less prediction theater, more evidence.
What an AI business actually sells
Not a model. A decision, a workflow, and a way to know when it stopped working.
School and paper
Master's, Business Analytics & AI
Ontario Tech University · 2025
3.99 / 4.3
A.I. Solutions Development
George Brown College · 2024
Dean's Honor List
Financial Technology
University of Toronto · 2022
Bachelor's, Health Sciences
University of Health Sciences, Pakistan · 2015
Also: IBM Data Engineering, IBM Full Stack, Azure Data Scientist Associate, EDHEC investment management with Python and ML. Useful stamps. Not the work.
Notes
Assessment call
A conversation before any teaching or advisory work. I use it to decide whether I can help. Duration and cost come after, not before.
Learn vs Sketches vs Contact
Learn is teaching and unblocking: a hard concept or a stuck project. Field notes are lessons from production. Sketches are those lessons, concepts, and prototypes shown visually. Contact is consulting and architecture for a small number of engagements.
Causal ML
A claim about what happens if you intervene, not a correlation you found in last year's prices. In pricing that means treating price as a treatment and refusing features that only exist because the price was already set.
Elasticity
How much demand moves when price moves, after you account for the things that jointly drive both. Businesses already have the word. A naive fit still sounds like common sense, which is the danger.
MLOps
The lifecycle that keeps a model from becoming folklore: a data contract, a training path that is not a laptop, a deploy you can roll back, monitoring on the decision, a retrain that is scheduled.
What this site is not
Not a tutorial farm, not a signal service, not a place that quotes before it has heard the work. If you want a certificate, look elsewhere.