TruEra AI Observability helps drive AI quality and performance, minimize LLM app hallucinations and risks
Redwood City, Calif. – Sept. 19, 2023 – TruEra today launched TruEra AI Observability, the first full-lifecycle AI observability solution providing monitoring, debugging, and testing for ML models in a single SaaS offering. TruEra AI Observability now covers both generative and traditional (discriminative) ML models, meeting customer needs for observability across their full portfolio of AI applications, as interest in developing and monitoring LLM-based apps is accelerating.
Minimizing the risks of LLM-based applications
Initial development of LLM-based applications is dramatically increasing since the launch of ChatGPT. However, LLM-based applications have well-known risks for hallucinations, toxicity and bias. TruEra AI Observability offers new capabilities for testing and tracking LLM apps in development and in live use, so that risks are minimized while accelerating LLM app development. The product capabilities were informed by the traction of TruLens – TruEra’s open source library for evaluating LLM applications.
“New applications and use cases based on Large Language Models like GPT-4, Google Palm, and open source models like Llama 2 from Meta are proliferating. A vast range of companies, from startups to major global enterprises, are inspired by the value that they could deliver,” said Tim Tully, Partner at Menlo Ventures and former CTO, Splunk. “Enthusiasm for LLMs is tempered by caution about the risks that LLM apps could potentially bring. TruEra AI Observability offers new LLM app testing, debugging, and monitoring capabilities that can help to minimize this risk.”
Full lifecycle AI Observability – now available as SaaS
Until the launch of TruEra AI Observability, observability tools generally focused on providing ML operations (MLOps) support for development or production, but not both. TruEra is the only software vendor providing a solution for driving high-performing, trustworthy AI across the full model lifecycle. TruEra, which previously only offered its full-lifecycle observability solutions either on-premises or virtual private cloud, now provides customers with the flexibility of SaaS or deploying within their own private cloud.
Optimizing AI models at AES
AES is deploying TruEra AI Observability after becoming an early adopter. AES, a global energy company, owns and operates significant numbers of wind and solar generating assets. The AES data science team has built cutting-edge models for wind and solar generation forecasting as well as for preventive maintenance.
“As AES expands the coverage and use of our AI models, it’s critical that we put in place MLOps tools to efficiently monitor, test, and debug our AI at scale,” said Sean Otto, Director of Analytics at AES. “TruEra’s powerful monitoring and testing capabilities enable us to increase the number and types of AI models that we use to optimize the management of our energy operations, which is critical in the dynamic industry and environmental conditions in which we operate. TruEra’s unique root cause analysis and explainability-powered analytics identify significant opportunities to improve model performance while partnering with our business teams. We view TruEra as a key part of our MLOps stack.”
Bringing powerful model monitoring, debugging, and testing capabilities to everyone
“TruEra’s initial success was driven by customers in banking, insurance, and other financial services, whose high security requirements were well met by existing TruEra on-prem solutions,” said TruEra Co-founder, President and Chief Scientist Anupam Datta. “Now, with TruEra AI Observability, we are bringing ML monitoring, debugging, and testing to a broader range of organizations, who prefer the rapid deployment, scalability, and flexibility of SaaS. We were excited to see hundreds of users sign up in the early beta period, while thousands have engaged with our hands-on educational offerings and community. The solution brings incredible monitoring and testing capabilities to everyone developing machine learning models and LLM applications.”
TruEra AI Observability helps data scientists and ML engineers to:
- Monitor models in production via customizable dashboards to ensure that they meet KPI and performance targets, as well as prevent failure scenarios such as model drift
- Debug quickly, using powerful feature and segment root cause analysis (RCA)
- Automatically run performance, quality and responsible AI tests to efficiently evaluate AI applications, validate improvements and prevent regressions
- Proactively identify predictive and generative AI application improvements using automated performance and quality analysis, RCA, AI explainability and model comparison
- Identify high-impact problems and then debug and optimize models quickly
A comprehensive approach to AI Observability – with powerful root cause analysis
Tru Era’s full lifecycle AI Observability approach ensures that teams are able to quickly identify emerging issues, isolate their root causes, and then debug and test quickly. For more on full lifecycle AI Observability, get the whitepaper – “How to Manage AI Performance – Full Lifecycle AI Observability vs. Production-Only and KPI Monitoring”
Free version of TruEra AI Observability now available
TruEra also introduced a free version of TruEra AI Observability focused on model testing, debugging, and evaluation. To try it, visit app.truera.net/ To learn more, visit truera.com.
About TruEra
TruEra is the leader in AI Observability, helping companies to monitor, debug, and test their ML models in order to drive performance, quality, and trustworthiness. Powered by enterprise-class Artificial Intelligence (AI) Explainability technology based on over nine years of research started at Carnegie Mellon University, TruEra is able to facilitate faster, more accurate ML model monitoring, analysis and debugging than any other vendor. Organizations using TruEra can achieve higher quality, higher performing models that sustainably achieve measurable business results, address unfair bias, and ensure governance and compliance.
