Tau
- Fully automated tabular learning engine, handing back a finished model out of the box
- Breakthrough accuracy, outperforming gradient boosting and tabular foundation models at every scale
- Handles real-world needs such as missing labels, random effects, selection bias and prediction intervals
Currently in Private Beta
Tau is deployed with a select group of customers. I am keeping access limited so I can work closely with them. Reach out to me to get onboarded.
What is Tau and why?
Tau is a fully automated learning engine for tabular data. Where a state-of-the-art model normally takes hand-built features, model selection and tuning, and validation at every step, Tau does all of it for you and hands back a finished model out of the box.
On benchmarks, Tau beats both gradient boosting and tabular foundation models, the two strongest approaches available today. It does so at every scale, from 30-record datasets to millions of records, while other engines are built for one end of that range and lose accuracy at the other.
Tau also handles messy, real-world data: missing labels, random effects, and selection bias. Every prediction also comes with a prediction interval. Projects often require it: manual analysis, experimental design, price optimization. Most machine learning models cannot provide these. In Tau all of this is native to the model - no clumsy workaround bolted on top.
Pricing
Research & Evaluation
FreeEvaluate Tau's capabilities against your own data and models.
- Full access to the modeling tool
- 500+ standard benchmark datasets
- Or benchmark against your own data
Consulting Engagement
IncludedEngage ArsChitectura to build and deliver models tailored to your business.
- Expert analysis and breakthroughs
- Tailored to your business context
- Full rights to use the deliverables
Direct License
Per Seat / MonthUse Tau independently for research, modeling, and production.
- Full access to the modeling tool
- Use outputs for any operational purpose
- Scale across your organization
Frequently Asked Questions
No matches.
What kinds of challenges is Tau built for?
To be completed.
How does Tau differ from gradient boosting?
To be completed.
Can I use Tau for quick diagnostics?
To be completed.
Does Tau work with scikit-learn pipelines and Pandas DataFrames?
To be completed.
What confidence intervals does Tau report?
To be completed.
Can Tau handle very small or very large datasets?
To be completed.
What Python versions and dependencies are required?
To be completed.
How do I install Tau?
To be completed.
What is Tau's license?
To be completed.
How can I get support?
To be completed.