About
A research studio for the data that runs organisations
TensorSpeed exists to make classification on structured data accurate, efficient and trustworthy — and to put that capability into the hands of the teams who make decisions with it.
Why tabular data
The overlooked majority of machine learning
Much of the attention in AI goes to text, images and video. Yet most decisions inside organisations — who to approve, what to inspect, which case to escalate — are made from rows and columns.
Classifying that data well is a hard, distinct problem: tables are heterogeneous, often small, frequently incomplete, and the cost of a confident mistake is real. It is the problem we have chosen to focus on.
Principles
How we work
Evidence over hype
Honest uncertainty
Built to be owned
Privacy by default
Research and practice
Two halves of one loop
Our research is shaped by the problems we see in practice, and our solutions use what the research teaches us. Findings that survive real data become part of how we build; failures in production become the next research question.
We share what we learn in our research notes and insights.
Have a classification problem worth solving?
Tell us about your data and the decision you need to make.