Snippets

Data platform engineer and AI-native builder — 10 years in data, 16 in engineering, Berkeley MIDS and P.Eng. Build production data platforms on Dagster, Snowflake, Polars, DuckDB and dbt, then ship AI-native systems on top of them: LLM-in-the-loop pipelines, agentic developer tooling with Claude Code and MCP, and structured-output services gated by evals. Rebuilt StackDX's US data platform as lead engineer and now run roadmap and code review for the 6-person team, still shipping weekly.

- Architected a production Dagster platform at StackDX across 15 US states and 945 transformation assets, distilling a 3.8 TB source layer into a 10.3 GB exposure layer serving 300M+ records to customer endpoints.
- Founded twochannel.ai: an AI-native catalog where LLM-in-the-loop pipelines and a custom Claude Code toolchain maintain 18k+ products across 2k+ brands — one engineer, end to end.
- Owned Canlin's regulated annual reserves evaluation for 5 years — the externally-audited valuation feeding the financial statements; technical revisions added +$730M (NPV10).
- Built the dbt Cloud / Snowflake platform that became Canlin's foundation for operational data, BI, and ML; automation from it saved 50+ hours/week across Operations.
- Set roadmap and review code for a 6-person, two-country data team at StackDX, while still shipping weekly in the platform codebases.

Moved from lead engineer into product ownership after rebuilding the US data platform: roadmap and code review for a 6-person team across two countries, while still shipping weekly in the Dagster and Stack Maps codebases.

Product ownership and team leadership:
- Run roadmap, prioritization, and code review for a 6-person team across US and Canadian data codebases — while still shipping in the Dagster platform weekly.
- Built PM tooling as software, in daily use: a Jira roadmap auto-scheduler, human-gated production-failure triage, and skills for weekly updates and Jira templating.
- Set Stack Maps product direction (frontend + API), scoping and prioritizing with the frontend and API engineers against the data-platform realities underneath.

Lead engineer on the USA data platform rebuild — replacing a legacy Python ETL stack with a production-grade Dagster-native architecture on AWS, with Polars and DuckDB doing the heavy transformation work. The platform distills a 3.8 TB source layer of 1.03M regulator files into a 10.3 GB exposure layer serving customer endpoints — 127M well production records, 172M lease production records, and 4.1M wellheaders. Built AI-first as well, on custom Claude Code skills, sub-agents and hooks covering scaffolding, data inspection, formatting and review.

Modern Data Platform Architecture:
- Architected a production Dagster platform across 15 US states and 945 assets, distilling a 3.8 TB source layer into a 10.3 GB exposure layer serving 300M+ records.
- Converted the Texas production fact build — the largest on the platform — to a streaming Polars plan, eliminating a repeat production out-of-memory failure.
- 146 contract-driven freshness checks across all 15 states, fanning out to 468 evaluations, threshold-calibrated to kill false-positive alerts.
- Designed reusable exposure/coalesce patterns so legacy mart schemas ride on the new fact-table architecture and new states extend the pattern instead of needing bespoke builds.
AI-native Developer Workflow:
- Built a project-specific Claude Code toolchain — a parquet-inspector sub-agent, an auto-format hook, and task skills for pipeline triage and schema work — over a token-optimization layer (618M tokens saved to date).
- Profiled my own Claude Code conversation history to find the highest-frequency ad-hoc work, then promoted the dominant pattern into a first-class sub-agent.
Cloud Infrastructure & DevOps:
- Deployed the AWS footprint with Terraform IaC across staging and production — RDS, S3, and ECS compute launched and managed per-job by Dagster.
- Implemented CI/CD (GitHub Actions, ruff, pyright, asset schema-contract tests) plus the branching and deployment automation behind staging → prod releases.
Framework & Team Enablement:
- Implemented 10 of 15 state pipelines solo, then onboarded two non-platform engineers who shipped the rest by following the documented conventions and schema contracts.

Designed and built Twochannel end-to-end — catalog, ingestion pipelines, search, recommendations, and frontend — rebuilt three times since 2021 as the stack evolved, publicly launched in 2026. The 18k+ product / 2k+ brand catalog is maintained almost entirely through AI-native workflows: LLM-in-the-loop scrapers, structured-output enrichment, semantic dedup, and a Claude Code toolchain operating against the full monorepo. The whole thing is the size that would normally need a small data team behind it, and it's just me.

AI-assisted product catalog at scale:
- Brand-scraper → LLM normalizer → reviewed dedup pipeline now maintains 18k+ products across 2k+ brands in Directus with minimal manual intervention.
- A 12-agent Claude Code curation team with human approval gates runs catalog curation — three pipeline defects caught at the gates, zero silent corruptions of production state.
- Wizard captures budget, room and genres; an async build job then returns complete system variants, constrained by signal-chain adjacency and per-category budget allocation.
- Replaced a three-tier dedup cascade and a fixed 0.9 confidence cutoff with one Claude decision service holding customer catalog writes to the curation team's standard.
Platform Architecture:
- Turborepo monorepo shipping Next.js (Vercel), Directus (Railway), Algolia search, Neo4j relationships, and Doppler-managed secrets across three hosting targets.
- Shipped GDPR Article 17 erasure — a twelve-step deletion pipeline with audit trail, CI-enforced FK rules, and weekly orphan sweep — backed by a 65-test Playwright E2E suite green on prod.

Senior member of the Data Analytics and Integration team — integrating best practices across the full data stack, mentoring ICs, and advocating for a modern data platform. Delivered polished, robust solutions spanning custom web apps, modern data-stack PoCs, data engineering pipelines, and high-fidelity dashboards. Toolkit: Databricks, PowerBI, Spotfire, dbt Core, cube.js, DuckDB; SQL, Python, R, DAX.

Dashboards & Corporate Reporting:
- Built the corporate PowerBI dashboard suite (Netback, Operating Cost, Capital) — corporate-to-well drill-downs used by executives, operations, production engineering, and development.
- Automated the weekly well production report end-to-end (Spotfire + Power Automate), removing a half-day weekly manual process from the analytics manager's plate.
Custom Web Applications:
- Replaced a manual 2–3-hour Excel frac-scheduling process with a Streamlit optimization app — its sequence plans ran ~8 multi-well pads.
Data Governance & Platform Advocacy:
- Local modern data stack PoC (Python / dbt / cube.js / DuckDB / PowerBI in Docker) replacing legacy R + CSV processes.
- Data-governance-committee advocate for cloud-first analytics; educated senior leadership on modern stack design.

Designed, built, and ran Canlin's modern data stack (dbt Cloud, Snowflake, DataRobot, Tableau, Streamlit) and led the Integrated Remote Operating Centre (IROC) analytics. Owned descriptive and predictive solutions end-to-end, mentored junior professionals into data roles, and made the business case to senior leadership for sustained investment in analytics.

Corporate Data Warehouse on dbt Cloud + Snowflake:
- Built the dbt Cloud / Snowflake platform that became Canlin's foundation for operational data, BI, and ML — SCADA to Accounting in a single source of truth.
- Automated the weekly production reporting process end-to-end — backend workflows, input sheets, and the report itself — saving 50+ hours/week across Operations.
Data Science & ML Ops:
- Deployed a modified XGBoost forecaster inferring daily across 8,000+ wells (30-day horizon), plus 2-year forecasts of the full well set, feeding corporate reporting.
- Built the real-time well-status analytics behind Canlin's remote operations centre (~500 wells), with networkx-based outage impact auto-assessment and anomaly-detection reporting.
- Streamlit SME-labeling app + classification model ranking well risk across the portfolio, with explanations.
- Home-grown ML-ops layer (Streamlit + Snowflake + NocoDB) for labeling and data input, with DataRobot for productionized models.

Drove Canlin's transition to a self-service data model by pairing data engineering with petroleum-data expertise — implementing Tableau as the company-wide reporting tool and data mart, advocating for repeatable BI, advanced analytics, and automation. Owned the regulated annual reserves evaluation — the externally-audited, company-wide valuation of every corporate asset, tied directly to the financial statements — improving accuracy and unlocking +$730M in value through technical revisions.

Tableau data-warehouse implementation:
- Implemented Tableau as the corporate reporting platform — 100+ data sources, 50 workbooks, and 20 Prep flows published in year one.
- Mentored every Tableau Creator/Explorer in the company and made the executive case for self-service data.
Corporate Reserves — regulated annual asset valuation, 5 years:
- Owned the externally-audited NI 51-101 reserves evaluation for 5 years — technical revisions added +$730M (NPV10) across four consecutive cycles.
Corporate Acquisitions & Divestitures:
- Built the data flows and dashboards for rapid A&D evaluation through a depressed-price divestiture cycle that brought corporate debt from $120M to $0

Exploitation/development engineering across multiple assets. Authored multi-scenario asset-longevity analysis that set 5–10-year corporate strategic direction.

Asset-longevity analysis & internal tooling:
- Multi-scenario asset-longevity modeling across all operated facilities that set 5–10-year corporate strategic direction
- Built a VBA-driven opportunity-tracking tool with a staging/approval workflow for development planning

On-site production engineering plus exploitation engineering across multiple assets. Built multi-dimensional tracking models from time-series sensor data and designed data-driven drill programs.

Highlights across both roles:
- On-site production/operations engineering, including long-term sensor-data trending and equipment monitoring
- Multi-dimensional tracking models built from time-series sensor (thermocouple) data
- Drill-program design and feasibility analysis across multiple assets

Multiple co-op terms as part of the University of Waterloo Engineering program — 8 months at Lanmark Engineering, 4 months at the Olds Sour Gas Plant, and 12 months in the Pengrowth office as an Exploitation Engineering co-op.