SaleCast - ML forecasting for multi-channel commerce
The challenge
A best-seller, a seasonal product and an end-of-life SKU do not follow the same curve. One forecasting algorithm cannot be honest for all of them.
What I built
SaleCast is the command center for a seller operating across many channels at once. It reconciles contradicting sources — for each field (price, stock, order) a single platform is the source of truth, so 'Amazon says 3, Shopify says 7' stops happening — and predicts what's coming (sales, demand, comms, traffic) with an engine where 12 algorithms compete per product. Under the hood it's a real multi-tenant SaaS: cross-tenant data access is blocked at compile time, and a zero-dependency .NET core powers four apps (Web, native Desktop, CLI, API).
The B2B buyer's first fear — can my data leak into another client's account? — is answered by architecture, not a promise. And ML is made observable, comparable and explainable enough that business users can trust or reject a forecast.
Key engineering points
Algorithm competition per SKU instead of one global forecast.
ML.NET and ONNX execution paths with targeted Python sidecars where foundation models make sense.
Drift detection to catch models that silently become wrong.
Price elasticity estimation with honest simulated examples while the product is still private.
Similar technical risk?
I can help scope the risk, architecture and first deliverable.
A 30-minute first call is enough to see whether I am the right profile for the problem.
A similar challenge?
A system like this one to build? Let's talk.
I take on critical technical work — from scoping to production, no debt or lock-in once it's handed over. Fastest way to see if it fits: a 30-minute call.
I reply within 24h — often sooner.