Braintrust lands $80M funding round to become the observability layer for AI – News

Artificial intelligence-native observability startup Braintrust Data Inc. believes it’s on a fast track to becoming an essentially infrastructure layer for AI models and agents in production after raising $80 million in a middle-stage funding today.

The Series B round was led by Iconiq and saw participation from high-profile backers, including Andreessen Horowitz, Greylock, basecase capital and Elad Gil, bringing Braintrust’s valuation to $800 million.

The startup has developed an AI-focused observability and evaluation platform that’s designed specifically for monitoring AI models and products. Unlike traditional monitoring systems that track elements such as system health, Braintrust provides a way to keep tabs on the quality of AI models and their outputs, and monitor them for hallucinations, drift or regression, which result in less accurate responses over time.

In a blog post, Braintrust co-founder and Chief Executive Ankur Goyal (pictured) explained why traditional observability tools don’t cut the mustard for AI-native companies and model designers. With AI tools moving rapidly from experimentation to production, agentic systems embedded in almost every engineering workflow, and models powering all manner of widely-used products, existing tools can’t keep up.

The problem is that there’s just too much for them to monitor, Goyal said. That includes traces longer than ever and multistep agents with tool calls and intermediate reasoning that create enormous volumes of data, amounting to hundreds of megabytes per interaction.

“Teams have never had less conviction about what will fail next,” Goyal said. “When something does break, it’s never been harder to explain why. Traditional observability can no longer keep pace with this scale, and teams are straining to juggle the demands of the business with the limitations of existing tools.

Braintrust changes this by integrating several important observability workflows, including exhaustive tracing tools that automatically capture every step of an AI model or agent’s reasoning process, including the prompts, tool calls, retrieved context and metadata on latency and cost. It also provides automated evaluation of models, using built-in scorers and a large language model-as-a-judge to evaluate their outputs for accuracy and relevance,. There’s also a playground to test and version-control prompt changes against real production data prior to deployment.

The startup has also developed an AI assistant to help companies observe their AI agents and models. By analyzing millions of traces, it cann suggest better prompts, create new datasets and identify patterns that could result in specific kinds of hallucinations, where models generate false answers. All of this runs on the company’s own, purpose-built database, called Brainstore, which it says is 80% faster at querying complex AI traces.

Goyal said Braintrust has won the loyalty of dozens of AI-native companies, including Notion Labs Inc., Replit Inc., Cloudflare Inc., Ramp Inc. and Dropbox Inc., which use its platform to monitor their AI deployments. One of the reasons for that is the close cooperation Braintrust provides. “I make sure to talk to [our customers] every day, listening to their difficulties and celebrating their wins,” Goyal said. “If Braintrust makes their lives easier and their products better, I know we are doing our job.”

That sentiment is shared by Iconiq General Partner Matt Jacobsen, who understands that customer success is key to driving the kind of growth that he wants to see in all of his portfolio companies. He’s in a position to know, since Iconiq had previously invested early in enterprise giants such as Salesforce Inc., Snowflake Inc. and Datadog Inc., as well as AI startups such as Anthropic PBC and the vector database firm Pinecone Systems Inc.

“We have seen that a defining trait among the generational companies is deep, authentic customer obsession,” Jacobsen said. “We believe Ankur and the Braintrust team embody this mindset and have been building their product from the start to serve the evolving needs of their customers.”

Goyal said Braintrust will use the funding to grow its engineering and go-to-market teams, launch new observability tools and expand into new regions.

Photo: Braintrust/YouTube