Analytics for 20,000 retail users at a tenth of the usual cloud cost

Context

alphashots.ai shows retail traders the historical base rate for a live price pattern: how often similar setups led to gains or losses, and the spread of risk and reward. It has to answer on demand, for non-experts, at a subscription price retail users will pay.

The problem

Searching the full price history of every traded stock for earlier occurrences of a pattern is compute-heavy. At standard cloud prices the product could not be sold at a retail price. And an overfitted base rate looks just as convincing to a user as a sound one, so the numbers had to be validated before anyone saw them.

What was built

• A matching engine that searches full price history and returns the empirical base rate and risk-reward distribution in under two minutes
• Infrastructure that runs the workload at about 10% of equivalent AWS cost
• Out-of-sample validation of every base rate before a pattern is shown to users