02Data platform · Generative AI
Weekly public data, parsed into a 52-week forecast the buying team takes into supplier negotiations.
Colombia's statistics office publishes wholesale food prices every week, in files made for people to read: layouts shift between releases, units change from sheet to sheet, and there is no API. Meanwhile the buying team negotiated with suppliers with no outside reference for what the market was doing.
The platform reads every release reliably, keeps years of history for hundreds of products, forecasts each price 52 weeks ahead, and has an AI model generate automated insights. It started as an internal tool and is now a secure web app the buyers take into supplier negotiations.
The tools, the architecture and the engineering numbers, one line each.
01MLOps · Time series
Over 100,000 SKU+Store combinations, forecast weekly by a routing policy that picks a model per cluster.
View full case study: ML demand forecasting in production03Applied AI · Guardrails
A Telegram assistant for questions about sales, inventory and purchasing. Every figure it gives comes from a query it actually ran.
View full case study: AI assistant for business data04Full stack · Internal product
Analysts used to work out every inter-store transfer by hand. The engine proposes what moves where, and they run it themselves.
View full case study: Stock rebalancing engine05Multi-agent · Local-first
The memory is markdown in Git, so it outlives the runtime that reads it.
View full case study: Personal AI operating system