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.
An agriculture technology company built a crop yield prediction model that combined satellite imagery, soil sensor data, weather forecasts, and historical yield records. The model predicted per-field
A pharmaceutical company I consulted for was three months into a project to fine-tune a large language model on their internal research corpus. The project had a team of four engineers, a budget of $8
A model registry is the version control system for your trained models. Without one, teams track model versions by filename, store artifacts in ad-hoc cloud storage locations, and discover which model
The regulatory landscape for AI safety has fractured along jurisdictional lines. The EU has taken a prescriptive, risk-based approach. The US has taken a sector-specific, agency-led approach. The UK h
Your office building has one electricity meter. At the end of the month, you get a bill for the whole building. You know the total cost of electricity for the month. You do not know which floor consum
When producers and consumers share a Kafka topic without agreeing on the data format, things break in production. A producer adds a field. A consumer expects the old schema. The deserialization fails,
The headline numbers are familiar. Women represent roughly a quarter of AI and data science professionals globally. At senior levels, the proportion drops to the low teens. At the C-suite level of AI-
A consumer goods company built an AI system that recommended reorder quantities for 12,000 SKUs across 340 distribution points. The system optimized for a multi-objective function that balanced invent
Code completion gets the attention, but it is the narrowest part of what AI can do in a development workflow. Walk into any team that has shipped software for a few years and they will tell you: writi