Writing.
Lessons from real projects on architecture, tools and leading engineers. New pieces twice a month. Leer en español
Check before you believe: what a perfect speed score leaves out
A headless demo against two real stores, 24 runs under identical conditions: why demos are fast, why one test misleads, and speed tricks with their costs.
SEO · October 8, 2026Best practices expire: keeping SEO, GEO and AEO current with AI
A dated log, a monthly AI-assisted review with a reusable prompt, and a test before every change.
Machine learning · October 4, 2026Kadabra: a Shazam for movies, and no way to know if it worked
Our 2023 Telegram bot that named movies from a photo or a plot: face recognition, age as a clue, vector search, and the evaluation it never had.
Data science · October 4, 202651,000 ads, 190 rows: revisiting a Kijiji visit-prediction model
Our 2023 model of Kijiji car-ad visits, the five problems that made its results meaningless, and a rebuild that beats the baseline by 41%.
Data science · October 4, 2026The threshold mattered more than the model: revisiting a bank marketing classifier
Our 2023 Kaggle model, seven bugs found three years later, and why tuning the F1 threshold beat every new model, feature and ensemble.
Delivery · September 25, 2026Finding delivery bottlenecks with lead time and cycle time
Lead time, cycle time, WIP and flow efficiency, five steps to find where work gets stuck, and the new review bottleneck AI creates.
Leadership · September 24, 2026How I measure engineering team health
Delivery, health and impact metrics, how they relate to DORA and SPACE, and how to tell whether AI assistants are really helping.
Leadership · September 22, 2026From data analyst to data engineer, through one good question
How one good question grew an engineer and improved our data pipelines, and where AI assistants fit in mentoring.
Architecture · September 21, 2026Pick the tool your team can support, not the one on the shortlist
What a six-times overrun on single sign-on taught us: a tool-selection scorecard, build vs buy for identity, and how AI assistants change the trade-off.
AI · March 5, 2024Shaping the future: what new graduates need for a career in AI
Updated for 2026: collaboration, real projects, the frameworks worth learning, working with AI assistants, ethics and lifelong learning.
AI · February 6, 2024Taming the data jungle: JSON as the contract between your systems and AI
Structured outputs, tool calling, MCP, memory and model configs, with a worked Pydantic example from real estate search.
Research · December 12, 2023Replacing random RT-PCR checks with data-driven traveller selection
Our STEM Fellowship paper: country risk levels from WHO data, age risk from Government of Canada data, and priority groups for travellers at Toronto Pearson.