Simor
Data Infrastructure for Production AI
Practical writing on AI data engineering, feature stores, and the infrastructure choices that determine whether AI systems work in production.
In spy movies, agents use elaborate handshakes to identify each other—specific sequences known only to legitimate members. One extends their hand a certain way, the other responds with the correct gri
Fine-tuning a 70B parameter model costs $50K+ and requires weeks of training on expensive hardware. This is the reality for teams building domain-specific language models. Traditional full-parameter f
Drop a rubber ball from shoulder height. It bounces back, but not as high. Each bounce is lower than the last—vigorous at first, then gradually settling, until it barely leaves the ground before final
Disney World on a summer morning. Thousands of families rushing toward gates. Without control, it would be a stampede. Enter the turnstiles: mechanical devices ensuring only one person passes at a tim
A food delivery platform's real-time dashboard froze during Friday dinner rush. Restaurants could not see incoming orders. Dispatchers could not assign drivers. Customer service was blind to delivery
On the Titanic, designers believed watertight bulkheads made it unsinkable. When the iceberg tore through multiple compartments, water spilled from one to another, creating a cascade that sank the "un
An insurance company's premium pricing model had been quietly going haywire for two weeks. Young drivers in high-risk areas were getting bargain prices while safe drivers faced astronomical quotes. By
Your home's electrical panel has circuit breakers. Plug in too many appliances, the breaker trips, cutting power to prevent fires. You can't use those outlets until you flip it back on. Annoying, but
An engineering team renamed a critical field from 'user_signup_date' to 'account_created_at' without warning. It cascaded through dozens of data pipelines, breaking executive dashboards, halting marke