Technical researcher · AI/product data engineer · founder of THE LINEUP
I build systems that have to survive contact with real data: API integrations, analytical pipelines, AI-assisted workflows, product instrumentation, and the checks that make their outputs inspectable.
My background began in journalism and moved through web development and data science into production AI engineering. That path still shapes how I work: define the question, preserve the sources, test the uncomfortable cases, state the limitations, and write the result so another person can reproduce it.
| Project | What it demonstrates |
|---|---|
| Job-board API benchmark | A reproducible comparison of public Greenhouse, Ashby, Lever, and SmartRecruiters endpoints, with timed runs, field-completeness measures, link checks, payload hashes, tests, and scheduled remeasurement. |
| Sports-data reliability lab | A synthetic, production-derived audit of provider coverage, freshness, canonical event identity, duplicate markets, and quote validity without exposing licensed data or proprietary product logic. |
| Evidence-first AI workflow | A deterministic control and evaluation layer for cited claims, missing or conflicting evidence, abstention, bounded retries, and human approval before external writes. |
| Evidence-grounded AI lead scanner | A public code example of an AI-assisted research workflow that keeps sources attached and a person responsible for the final decision. |
| Game Theory analysis | A recent Python research project built around long-form podcast analysis. |
| THE LINEUP case study | The architecture and operating evidence behind a live sports analytics product spanning provider data, models, market comparison, web/iOS delivery, subscriptions, monitoring, and automated grading. Production code remains private by design. |
- Reproducible technical research and honest benchmarks
- API and provider reliability, freshness, and source lineage
- AI workflows with evidence, abstention, evaluation, and human approval
- Product analytics that connect system behaviour to customer outcomes
- Clear writing about what worked, what failed, and what is still uncertain
The repositories behind THE LINEUP and my consulting systems are private because they contain proprietary product logic, infrastructure details, customer workflows, and licensed data integrations. I publish clean-room examples, synthetic fixtures, test rigs, and technical case studies instead of exposing production history or pretending a demo is the live system.
More work and case studies: duncananderson.ca
Contact: duncan@duncananderson.ca · Montreal, Canada