NeuBird Founders Leverage Cloud-Native Expertise to Innovate with Generative AI
Goutham Rao and Vinod Jayaraman, the founders of NeuBird, previously developed Portworx, a cloud-native storage solution that was sold to PureStorage in 2019 for an impressive $370 million—marking their third successful exit.
Upon searching for their next entrepreneurial venture last year, they recognized a prime opportunity to blend their deep expertise in cloud-native technologies, particularly in IT operations, with the rapidly evolving field of generative AI.
Today, NeuBird announced it has secured a substantial $22 million investment from Mayfield, a significant sum for an early-stage startup. This backing likely stems from confidence in the founders’ proven track record in building successful companies.
As CEO Rao points out, the cloud-native community has effectively addressed numerous complex problems over the last decade. However, this progress has introduced heightened complexity. "Our community has excelled in creating sophisticated cloud-native architectures with service-oriented designs. While beneficial, this has led to multiple layers, which also means increased telemetry and complexity," Rao explained.
They observed that the overwhelming amount of data generated made it nearly impossible for engineers within large organizations to effectively identify, diagnose, and resolve issues at scale. Concurrently, as large language models began to advance, the founders decided to harness this technology to tackle these challenges.
"We're uniquely applying large language models to analyze extensive metrics, alerts, logs, traces, and application configurations in mere seconds. This lets us assess the health of the environment, detect issues, and propose solutions,” Rao stated.
Essentially, the company is creating a reliable digital assistant for engineering teams. "It's a digital co-worker that collaborates with Site Reliability Engineers (SREs) and IT Operations engineers, monitoring alerts and logs for potential issues," he added. Their aim is to drastically reduce incident response and resolution times from hours to mere minutes. By leveraging generative AI, they believe they can significantly enhance companies' operational efficiencies.
The founders are acutely aware of the limitations inherent in large language models. To combat inaccuracies, they will utilize a focused dataset for model training and implement additional systems that ensure the precision of responses. "By controlling the data used for very specific environments, we can validate the AI's outputs against a vector database, ensuring accuracy. If we're unsatisfied with the results, we won't recommend them to users," he affirmed.
NeuBird allows customers to connect directly to their cloud systems by entering their credentials. This approach lets NeuBird access and cross-check available data without transferring it, simplifying the process of obtaining company-specific information for effective model training.
The startup employs various models, including Llama 2 for log and metric analysis and Mistral for different analytical tasks. They effectively convert every natural language interaction into a SQL query, transforming unstructured data into structured formats, which they anticipate will enhance accuracy considerably.
Currently, NeuBird is collaborating with design and alpha partners to refine their offering, aiming for a market launch later this year. Rao emphasized the significance of their initial funding, noting that it provides the necessary latitude to focus on solving the problem without the immediate pressure of seeking additional investment.
Ranking
