mnelson.ca

I build AI-native data platforms — and multiply the teams that run them.

Matt Nelson · P.Eng · Berkeley MIDS · 10 years in data, 16 in engineering · Calgary, Canada

The path so far

  1. 2005–2010

    University of WaterlooBASc Chemical Engineering

  2. 2006–2010

    Engineering Co-opLanmark · Olds Gas Plant · Pengrowth

  3. 2010–2014

    Pengrowth EnergyProduction & Exploitation Engineer

  4. 2014–2017

    Centrica EnergyExploitation / Development Engineer

  5. 2016–2018

    Berkeley MIDSMaster of Information & Data Science

    Master of Information and Data Science at UC Berkeley (3.89 GPA) — completed nights-and-weekends while working full-time. The pivot point where a decade of engineering turned into a data career.

  6. 2017–2022

    Canlin EnergyCorporate Data Scientist & Reserves Manager

    Drove the transition to self-service data: implemented Tableau company-wide (100+ data sources, 50 workbooks in year one) and owned the regulated annual reserves evaluation feeding audited financial statements.

  7. 2022–2024

    Canlin EnergyData & Advanced Analytics Lead

    Designed, built, and ran Canlin’s modern data stack (dbt Cloud, Snowflake, DataRobot) and led IROC (remote operations centre) analytics — mentoring junior professionals into data roles along the way.

  8. 2024–2025

    Paramount ResourcesSenior Data Analytics Engineer

    Senior IC on the Data Analytics and Integration team — corporate PowerBI dashboard suites, Streamlit optimization apps, and governance-level advocacy for a modern cloud data stack.

  9. 2025–2026

    StackDXData Platform Engineer

    Lead engineer on the US data platform rebuild: a legacy Python ETL stack replaced with a production Dagster architecture — 15 states, 945 transformation assets, a 3.8 TB source layer distilled to a 10.3 GB exposure layer serving 300M+ records — shipped AI-first with a custom Claude Code toolchain. Read more

  10. 2026–now

    StackDXProduct Manager, Stack Maps & Public Data

    Product Manager for Stack Maps and the Public Data team — a 6-person group across two countries — setting roadmap while still shipping in the codebases the roadmap depends on.

Flagship projects

  1. 2014

    P.Eng designationAPEGA, 2014

    Professional Engineer designation (APEGA) — the licensed-engineering rigor that underwrites every production system since.

  2. 2017–2022

    Reserves evaluation ownership+$730M technical revisions

    Owned the externally-audited NI 51-101 reserves evaluation for five years — the board-level valuation of every corporate asset. Rigorous data and engineering work added +$730M (NPV10) in technical revisions. Read more

  3. 2021–now

    twochannel.ai18k+ products · one engineer

    Founder project: an AI-assisted HiFi catalog and system designer where LLM-in-the-loop pipelines maintain 18k+ products across 2k+ brands. The catalog is the size that would normally need a small data team behind it, and there is only me. Read more

  4. 2022–2024

    dbt + Snowflake platformCanlin’s data foundation

    The dbt Cloud / Snowflake platform that became Canlin’s foundation for operational data, BI, and ML — automation from it saved 50+ hours per week across Operations. Read more

  5. 2022–2024

    8,000-well forecastingdaily production ML

    Modified XGBoost models forecasting the next 30 days of production for 8,000+ wells, inferring daily into corporate reporting — plus 2-year forecasts across the full well set. Read more

  6. 2022–2024

    IROC analyticsremote operations centre

    Real-time well status and production dashboards for Canlin’s Integrated Remote Operating Centre, with facility-outage impact auto-assessment built on networkx.

  7. 2023–now

    Claude Code toolchainskills · sub-agents · hooks · MCP

    A cross-repo AI-native development practice: custom skills, sub-agents, hooks, and MCP servers — including profiling my own agent usage to find where the time was actually going, then promoting the dominant pattern into a first-class sub-agent.

  8. 2024–2025

    Auto-Frac sequenceroptimization app

    Streamlit app replacing a manual operations-scheduling task with a purpose-built optimization algorithm balancing efficiency, sequencing, and equipment-rotation constraints.

  9. 2024–2025

    Modern-stack PoCPython · dbt · cube.js · DuckDB

    A fully-working local modern data stack (Python ingestion → dbt Core → cube.js → DuckDB → PowerBI, in Docker) that replaced legacy R + CSV flows and anchored the case for cloud analytics.

  10. 2025–now

    US Dagster platform300M+ records · 945 assets

    Production-grade Dagster-native platform for US oil & gas data: Polars/DuckDB performance layer, Terraform-managed AWS, schema contracts, and CI — the case study in shipping a modern data platform AI-first. Read more

What it adds up to

  • Data platform engineering

    Dagster, dbt, Snowflake, Polars, DuckDB — three generations of platform builds, each one the foundation its company still runs on.

  • ML & AI engineering

    LLM-in-the-loop pipelines, structured outputs, MCP servers and Claude Code toolchains on the AI side, and forecasting, classification and anomaly detection actually deployed on the ML side — 8,000 wells inferring daily into corporate reporting, not notebooks.

  • Analytics engineering & BI

    Tableau and PowerBI as corporate platforms rather than one-off dashboards, semantic layers over the warehouse, and reporting automated end to end — the weekly production report alone gave Operations back 50+ hours a week.

  • Data leadership

    Team coordination, mentorship, and board-level ownership — currently a 6-person cross-border team, with a decade of turning analysts into data professionals behind it.

Live preview of Twochannel's system builder — a flagship AI-native monorepo designed and built end-to-end.

Hiring for a data or AI team?

The fastest way to reach me is email — or the resume is one click away.