Truera Launches Industry’s First Model Intelligence Platform, With Funding Led by Greylock, to Solve Machine Learning’s Black Box Problem

Redwood City, Calif. — August 3, 2020 — Truera, a startup focused on making machine learning models understandable, has emerged from stealth with what it calls a Model Intelligence platform. The software is designed to lift the “black box” surrounding machine learning and give teams insight into their models across the full lifecycle, from development through to production. The company also disclosed a $5.1 million first funding round led by Greylock.

Wing VC, Conversion Capital and angel investor Aaref Hilaly also took part in the round. Truera said its platform is already in use at early customers among the Fortune 100, particularly in banking and insurance, two industries where model decisions face heavy regulatory and reputational scrutiny.

The black box problem

Machine learning systems are prized for their ability to find complex patterns in data, but that same complexity makes them hard to interpret. In many cases, even the data scientists who built a model cannot fully explain why it produced a particular output. That opacity raises questions about trust, fairness, bias and transparency, and it can slow adoption in sectors where decisions about credit, insurance or customers must be justified.

A second challenge emerges after deployment. Models are trained on historical data, and when real-world conditions shift, live inputs can diverge from what the model learned, a phenomenon known as concept drift. The COVID-19 pandemic has made the issue especially visible, as abrupt changes in consumer and economic behavior have undermined assumptions baked into many production models.

What the platform does

Truera’s platform is built on its AI.Q technology, which uses sensitivity analysis to explain why a model reaches a given prediction. The company says it can be deployed on-premises or in a private cloud within hours, and it is aimed at four broad needs:

• Explaining and analyzing how machine learning models behave.
• Improving model quality during development.
• Building trust with stakeholders and reducing project risk.
• Monitoring production models for drift between training data and operational data.

Standard Chartered was named as an early adopter. The bank’s retail banking leadership said it wants to keep using data and AI in ways that remain fair, transparent and responsible, a goal that explainability tools are meant to support.

Truera is pitching the product at the whole team behind a model, not just its builders. Data scientists can use it to diagnose why a model underperforms, risk and compliance staff can use it to review how decisions are reached, and business owners can use it to judge whether a model is ready for production. By putting explanation and monitoring on one platform, the company aims to replace the patchwork of ad hoc scripts and manual reviews many organizations still rely on to validate their models.

Rooted in Carnegie Mellon research

Truera was founded by Will Uppington, Anupam Datta and Shayak Sen, who teamed up in 2019. Datta, a Carnegie Mellon University professor, and Sen, who earned his PhD there, spent about six years developing AI explainability technology at the university, work that underpins the company’s product. Uppington, who serves as CEO, previously led product and customer-facing teams at another startup, where he saw firsthand how a lack of insight into models hampered development, sales and customer success.

Uppington said the company’s aim is to build software that “helps every data scientist analyze, improve and build trust in their models.” Greylock partner Jerry Chen, whose firm led the round, argued that solving the black box problem is essential for building effective and responsible machine learning applications.

Why it matters

Truera’s debut comes as enterprises move machine learning from experiments into decisions that directly affect customers. As that happens, the ability to explain, audit and monitor models is shifting from a research topic to an operational requirement, driven by regulators, risk teams and growing public concern over algorithmic bias.

The pandemic has added urgency by exposing how quickly models can degrade when the world changes. Tools that combine explainability with ongoing drift monitoring address both the trust problem and the reliability problem at once. Truera enters a nascent but increasingly crowded field of explainable AI and model monitoring vendors; its academic pedigree, backing from Greylock and early traction with large financial institutions give it a credible starting position as demand for responsible AI tooling grows.