Outerbounds, a machine learning infrastructure startup, today announced new product capabilities to help enterprises prepare for and adopt generative AI models like ChatGPT.
The company’s cofounders, CEO Ville Tuulos and CTO Savin Goyal, both former Netflix data scientists, aim to position Outerbounds as a leading provider of ML infrastructure as businesses increasingly look to leverage large language models (LLMs).
The new features added to the platform include GPU compute for generative AI use cases, bank-grade security and compliance, and workstation support for data scientists. These features aim to help customers ship data, ML, and AI projects faster, while retaining control over their data and models.
Tuulos explained the rationale of the new features in a recent interview with VentureBeat, stating, “The adoption of generative AI and LLMs should not be a quick fix or a gimmick. It should be tailored to enhance a company’s products in meaningful ways.”
“Although AI is new and shiny and exciting today, in the long term AI isn’t an excuse to provide a subpar product experience,” he added. “The best companies will learn how to adapt and customize AI techniques to support their products in specific ways, not just as an easy chat add-on.”
Since the startup launched in 2021, Outerbounds has been instrumental in the success of several businesses such as Trade Republic, Convoy and Wadhwani AI. Notably, Trade Republic deployed a new ML-powered feature in just six weeks, leading to a direct uplift in product metrics, thanks to Outerbounds.
Outerbounds is built on Metaflow, an open-source framework that the founders of Outerbounds created at Netflix in 2019. Metaflow is currently used by hundreds of leading ML and data science organizations across industries, such as Netflix, Zillow, 23andMe, CNN Media Group and Dyson.
Tuulos said that Outerbounds has added a unique approach to MLOps and managing the ML lifecycle, one focused on the user experience rather than technical capabilities.
“Ever since the beginning, we have focused on the user experience,” Tuulos said. “Since the field is so new, many other solutions have focused on technical capabilities, with the UX as an afterthought. We have always believed that the technology will mature, and as always, ultimately it is the best user experience that wins.”
Despite the complexities of AI and ML, Outerbounds has been able to use its experience to navigate the immature and chaotic landscape. “Having a solid foundation for any AI project is critical,” said Tuulos, highlighting the need for data, compute, orchestration and versioning in any AI project.
Outerbounds cofounder and CTO, Savin Goyal, echoed Tuulos’s sentiments on the importance of building a solid AI foundation. He said, “ML and AI should meet the same security standards as all other infrastructure, if not more.”
“We follow a cloud-prem deployment model,” Goyal added. “Everything runs on the customer’s cloud account with their own security policies and governance. We integrate with Snowflake, Databricks and open-source solutions.”
Goyal also said that Outerbounds helps customers address challenges like model governance, transparency and bias that come with deploying generative AI models.
“Our view is that there can’t be — and there shouldn’t be — a single entity dictating what bias means and what is acceptable when it comes to gen AI. Each company should be responsible for these choices based on their understanding of the market — similar to how companies are responsible for their behavior today even without gen AI,” he said. “We give companies tools so they can customize and fine-tune gen AI to their own needs.”
Outerbounds stands out in a crowded market with a unique approach to ML operations. “We are building a human-centric infrastructure that makes data scientists and data developers as productive as possible,” said Tuulos.
With the feature update, Outerbounds aims to solve the problem of data access, which Goyal sees as a “fundamental bottleneck.” He said, “How much time does it take for an individual to iterate through a variety of different iterations and hypotheses? If you’re spending 20 minutes to access the data that you need, it naturally breaks your flow state.”
The features released today further align Outerbounds with its mission to make it easier for companies to adopt ML and AI in more parts of their business. The company envisages a future where AI and ML can be applied everywhere, and these new enhancements are a step towards realizing this vision.
As the field of AI continues to evolve, businesses are grappling with the complexities of implementation and governance. Outerbounds, with its new features, is positioning itself at the forefront of this transformation, offering solutions that are not only technologically sophisticated but also mindful of user experience and governance concerns. With its new offerings, Outerbounds is paving the way for broader and more effective use of AI and ML in the enterprise.
Outerbounds, a machine learning infrastructure startup, today announced new product capabilities to help enterprises prepare for and adopt generative AI models like ChatGPT.
The company’s cofounders, CEO Ville Tuulos and CTO Savin Goyal, both former Netflix data scientists, aim to position Outerbounds as a leading provider of ML infrastructure as businesses increasingly look to leverage large language models (LLMs).
The new features added to the platform include GPU compute for generative AI use cases, bank-grade security and compliance, and workstation support for data scientists. These features aim to help customers ship data, ML, and AI projects faster, while retaining control over their data and models.
Tuulos explained the rationale of the new features in a recent interview with VentureBeat, stating, “The adoption of generative AI and LLMs should not be a quick fix or a gimmick. It should be tailored to enhance a company’s products in meaningful ways.”
“Although AI is new and shiny and exciting today, in the long term AI isn’t an excuse to provide a subpar product experience,” he added. “The best companies will learn how to adapt and customize AI techniques to support their products in specific ways, not just as an easy chat add-on.”
Leveraging its Netflix roots
Since the startup launched in 2021, Outerbounds has been instrumental in the success of several businesses such as Trade Republic, Convoy and Wadhwani AI. Notably, Trade Republic deployed a new ML-powered feature in just six weeks, leading to a direct uplift in product metrics, thanks to Outerbounds.
Outerbounds is built on Metaflow, an open-source framework that the founders of Outerbounds created at Netflix in 2019. Metaflow is currently used by hundreds of leading ML and data science organizations across industries, such as Netflix, Zillow, 23andMe, CNN Media Group and Dyson.
Tuulos said that Outerbounds has added a unique approach to MLOps and managing the ML lifecycle, one focused on the user experience rather than technical capabilities.
“Ever since the beginning, we have focused on the user experience,” Tuulos said. “Since the field is so new, many other solutions have focused on technical capabilities, with the UX as an afterthought. We have always believed that the technology will mature, and as always, ultimately it is the best user experience that wins.”
Seamless integration and bank-grade security
Despite the complexities of AI and ML, Outerbounds has been able to use its experience to navigate the immature and chaotic landscape. “Having a solid foundation for any AI project is critical,” said Tuulos, highlighting the need for data, compute, orchestration and versioning in any AI project.
Outerbounds cofounder and CTO, Savin Goyal, echoed Tuulos’s sentiments on the importance of building a solid AI foundation. He said, “ML and AI should meet the same security standards as all other infrastructure, if not more.”
“We follow a cloud-prem deployment model,” Goyal added. “Everything runs on the customer’s cloud account with their own security policies and governance. We integrate with Snowflake, Databricks and open-source solutions.”
Goyal also said that Outerbounds helps customers address challenges like model governance, transparency and bias that come with deploying generative AI models.
“Our view is that there can’t be — and there shouldn’t be — a single entity dictating what bias means and what is acceptable when it comes to gen AI. Each company should be responsible for these choices based on their understanding of the market — similar to how companies are responsible for their behavior today even without gen AI,” he said. “We give companies tools so they can customize and fine-tune gen AI to their own needs.”
Human-centric approach to ML operations
Outerbounds stands out in a crowded market with a unique approach to ML operations. “We are building a human-centric infrastructure that makes data scientists and data developers as productive as possible,” said Tuulos.
With the feature update, Outerbounds aims to solve the problem of data access, which Goyal sees as a “fundamental bottleneck.” He said, “How much time does it take for an individual to iterate through a variety of different iterations and hypotheses? If you’re spending 20 minutes to access the data that you need, it naturally breaks your flow state.”
The features released today further align Outerbounds with its mission to make it easier for companies to adopt ML and AI in more parts of their business. The company envisages a future where AI and ML can be applied everywhere, and these new enhancements are a step towards realizing this vision.
As the field of AI continues to evolve, businesses are grappling with the complexities of implementation and governance. Outerbounds, with its new features, is positioning itself at the forefront of this transformation, offering solutions that are not only technologically sophisticated but also mindful of user experience and governance concerns. With its new offerings, Outerbounds is paving the way for broader and more effective use of AI and ML in the enterprise.
Author: Michael Nuñez
Source: Venturebeat
