Domain experts’ workflow and data science in the R&D, production, and manufacturing world today fits hand in glove. If you’re a technologist, working with data and continuous exploration and monitoring are commonplace, but supplementing their analysis with cutting edge AI models isn’t. In an effort to patch the skill set gap hindering organizations from tapping into AI, Israeli startup Vanti Analytics has developed a self-service data science platform to equip technologists with universal predictive AI models at a fraction of the cost of prevailing methods.
“We’re creating technology intelligence, as opposed to business intelligence, to help technologists with data science tools and leverage them in their daily work,” said Vanti Analytics’ CTO and co-founder Niro Osiroff, who together with co-founder and CEO, Smadar David, previously led large-scale MEMS and algorithm groups at Israeli LiDAR startup Innoviz Technologies (both first employees) and the Israel Defense Forces’ 81 intelligence unit.
Vanti’s AI-as-a-service platform is designed to uniquely automate data science services for non-data science experts in manufacturing, production and decision-making industries. Through a highly intuitive and code-free platform, Vanti’s platform delivers actionable insights for any researcher, analyst, engineer or scientist relying on iterative data and scientific experimentation across the life-cycle production of products. “We want to automate their data science process but without the need to bring onboard a data science team or engage with an AI consulting company,” both at a steep price explained Smadar.
Using machine learning algorithms and novel system architectures, Vanti’s platform can help non-data science experts in various use cases and enable the implementation of production-level predictive and maintenance models. And with their ‘white-box’ neural network architecture – explainable AI – users can understand their experiment’s insights in contextual natural language.
The startup is targeting technologists whose daily working nature relies on the notion of scientific experimentation, like that of semiconductors, electronics manufacturing, life sciences and biomedical companies. “When you’re trying to integrate components together or develop a new drug, the process is to iterate the successful experiments and then race to the market with a product, ” explained Smadar. The ensuing iterations, ramp-ups and production monitoring creates new data points (think new machine, test or experiment logs) which lends itself well to a data science expert or team that’s focused on bringing a product to the market with the best performance and lowest cost. Although accessing those insights today is a premium good.
“The cost of data science today is astronomical,” said Niro. “If you want to beef up your data science team, it’s often at the expense of displacing other employees in the organization, and the team is usually dedicated to one project.” Aside from the rising salaries associated with hiring AI staff, or outsourcing to an AI consultancy, the cost of data acquisition can easily bear the same weight. “For example, obtaining another blood sample from a patient for a biomedical company could be considered only a mere data point, but the time and cost extrapolated to a scaled project can be astronomical.”
Vanti’s platform’s architecture is built from the ground up, for technologists, whether it’s understanding decision trees, formulas, or the datasets features and inherent biases. In three simple steps, users ask a question, connect their dataset, and receive insights from the platform’s engine. Vanti’s platform’s pretrained models are suited for myriad KPIs in the R&D and manufacturing sectors. It expedites data collection and helps identify root causes for observed phenomena, and it can support the manufacturing and testing aspect in the lifecycle of products to identify factors affecting yield and performance, and recommends improvements for the next iteration. Smadar likened their platform to “having a data scientist sitting next you.” Additionally, their algorithms are designed to distill simple English insights from a model, where users can ask follow up questions or investigate models, providing full visibility into its inner-workings.
“Our model is a ‘white-box’ model; providing a very explainable and semantically understood context of the domain,” said Niro. “Technologists want a model they can work with and trust to understand and improve for the next iteration. It’s a feature we believe will help companies overcome hesitation in adopting AI in the R&D and manufacturing process, where you actually care what’s going on ‘under-the-hood’ and where trust is paramount.”
By way of explainable AI, Vanti’s platform allows technologists to act on generated insights, for instance, ‘which parameter is the main contributor to the result’, or that by ‘adding 100 measurements of false parameter would improve accuracy by 98%’.
For a chip integration company deploying an ASIC for 5G network, Vanti’s platform was able to identify the most impairing parameters and recommended corrective measures. For a chip design and validation company, Vanti’s platform was able to enhance the Netlist Simulations with a closed-form analytic formula, reducing the simulation process from weeks to minutes. And for a medical device manufacturer, Vanti was able to make recommendations on future data points to collect based on predicted performance of generated design alternatives, thereby reducing the number of tests required to be executed.
The startup soft-launched their platform earlier in 2019, on an invite-only basis, and conducted several POCs, of which several have become paying clients. The subscription based SaaS platform is priced on compute usage for individual and enterprise versions.
Vanti Analytics raised a pre-seed round several months ago from i3 Equity Partners. They’re currently focused on product development and acting on insights generated by their clients.
In a recent Gartner survey of business executives, it found that AI implementation grew by 270% from 2016 to 2019. Indeed, AI technology is invariably penetrating real world cases, but the shortage of data science experts isn’t abating anytime soon. For the technologists that aren’t afforded access to data science teams and outsourcing experts, Vanti’s platform could become a decisive resource for many, leveling the playing field of AI.
Click here to apply for access to Vanti’s private beta version.