Real systems, real numbers
Every engagement below shipped to production and is still running. The metrics are measured post-launch, not projected before it.
Nimbus Retail cut stockouts by a third in one quarter
Nimbus was forecasting inventory across 40 stores in a single shared spreadsheet, updated by hand every Monday. Stock either sat unsold in the wrong warehouse or ran out on the shelf entirely — both cost real revenue every week.
We replaced it with a demand-forecasting model trained on two years of POS history, weather data and regional holidays, wired directly into their warehouse management system. Store managers now see a 14-day forecast refreshed nightly, with automatic reorder suggestions they can approve in one tap.
Filter by industry
Fieldbase — Data Platform Rebuild
A unified warehouse replacing six disconnected internal tools, handed over fully documented.
Kanpai Foods — Recommendation Engine
A real-time personalization layer added to an existing e-commerce checkout flow.
Loomstack — Anomaly Monitoring
Production monitoring with automatic anomaly detection replacing a set of brittle manual dashboards.
Torque Components — Visual QA
A computer-vision defect detector installed on an existing production line, no hardware replacement required.
Fielo — Conversational Support Agent
An LLM-based triage and resolution agent handling first-line support conversations end to end.
Nimbus Retail — Demand Forecasting
A nightly-refreshed forecasting model replaced a manual spreadsheet process across 40 stores.