TruEra’s technology helps evaluate the quality of input and output of large language models (LLMs).
With this, Snowflake aims to address the need for AI applications to actually be suitable for real-world use and continue to perform highly in production. “Not only must enterprises ensure accurate, reliable, and valuable results they must also address and mitigate critical issues like bias, hallucinations, and toxicity,” Snowflake said. According to the company, demonstrating that AI is trustworthy and high performing is critical to acceptance.
Tip: Snowflake enters the LLM war with introduction of Arctic
Data Cloud expansion
After previous introductions to advance the development and use of AI, Snowflake is now going to further invest in quality improvement with the acquisition of TruEra. As a result, the Data Cloud may gain new features for evaluating, monitoring and debugging models and apps. Users can do this in the development phase, as well as when the models and apps are already running in production.
Snowflake previously provided the Data Cloud with data governance capabilities, which should improve the accuracy and reliability of data used to train models.
TruEra’s observability options promise to complement that. In particular, TruEra should help at the LLM level by bringing experiment evaluation to applications such as question answering, summaries, RAG-based applications and agent-based applications. The technology can also provide detailed insights to improve model performance and accuracy by identifying anomalies.
It is not known exactly how much Snowflake is paying for TruEra. TruEra’s staff and the three founders, Anupam Datta, Shayak Sen and Will Uppington, will stay on to bring observability capabilities to the Data Cloud.
Snowflake growth
Snowflake announced the acquisition of TruEra simultaneously with the presentation of its quarterly results. Those showed that the company continues to grow for the time being. Year-on-year, the company saw its revenue increase by 33 percent to $829 million (€765.5 million), exceeding market expectations of $788 million. Consequently, after hours, Snowflake shares rose more than 4 percent. Net loss, on the other hand, was less rosy, with a $317 million loss.