This post was originally published on this site
AI continues to release a lot of new models and companies are racing to adopt them. According to Searce, nearly 10% of companies plan to invest a whopping 25 million dollars this year in AI initiatives.
While AI is a big part of the economy, it is unclear whether or not there is a return on investment. Gartner reports that half of AI leaders don’t know how to calculate or prove the value of AI project.
Chetan Sharma, an ex-Airbnb data science expert, argues that calculating AI ROI can be done with the right tools. Sharma is a co-founder of Eppo, an experimentation platform which allows customers to evaluate and customize AI models according to specific use cases. Eppo offers a general A/B-testing platform and service, in addition to its model evaluation suite.
Sharma told TechCrunch that A/B testing is a cost-effective method to evaluate the effectiveness of AI models. “Eppo helps businesses identify which models deliver value and allows smarter, more sustainable decision in an environment of rapid innovations and escalating cost.”
Eppo competes with a number of experimentation and A/B testing startups in the market, including <a href="https://techcrunch.com/2018/02/22/split-series-b/#! Eppo competes with a number of experimentation and A/B testing startups in the market, including a href="https://techcrunch.com/2018/02/22/split-series-b/#! AWS, Microsoft Azure, and Google Cloud are among the many tech giants that offer model evaluation and fine-tuning tools.
Sharma says Eppo is unique because of features such as its “contextual-bandit” system. This system automatically identifies new variants of customer websites, apps, or AI models, and actively explores their performance by increasing traffic or load.
Sharma said that “Experimentation accelerates growth and increases velocity by removing bureaucratic — often incorrect — committee decisions while tightly tying initiatives to growth metrics. This kills bad ideas quickly while canonizing the good ideas for reinvestment.” “Eppo’s approach to live ‘online-eval’ tests of AI models reveals whether premium models improve metrics.”
Sharma says that Eppo, ‘s stealth launch in 2022 has “several hundreds” of enterprise customers, including Twitch and SurveyMonkey. It also includes DraftKings Coinbase Descript Perplexity. Alexis Weill is Perplexity’s data head. He told TechCrunch Eppo had allowed Perplexity “significantly scale up” the number experiments it runs simultaneously.
Investors appear to be pleased. This week, Eppo completed a Series B round of $28 million led by Innovation Endeavors, with participation from Icon Ventures Amplify Partners, and Menlo Ventures. Sharma says the new cash will be used to bolster Eppo’s AI experimentation and marketing capabilities, enhance its analytics offerings, and scale its go-to market efforts.
Eppo, a San Francisco-based company, currently employs 45 people and expects to have 65 by the end of the year.
Sharma said that the combination of AI and the demands for efficient growth has created a mentality of adapt or die, which forces companies to be experimental. “And because of the gaps in legacy vendors, the majority of the experimentation market chose to staff large teams and build rather than buy. These in-house teams can no longer be sustained with so many employee movements and layoffs. This leads to companies looking to Eppo as a replacement for expensive or orphaned builds.