Vectice raises $12.6M to help enterprises document their data science assets

Spurred by its revenue-boosting potential, companies are increasingly embracing AI technologies across their organizations. Harris Poll, working with Appen, found that 55% of businesses accelerated their AI strategies in 2020 due to the pandemic. But a data science skills gap threatens to stymie progress. In a recent O’Reilly report, a lack of skilled people topped the list of challenges in AI, with 19% of respondents citing it as a “significant” barrier.

The skills gap isn’t the only barrier standing in the way of AI deployment. Data scientists at companies don’t always have sufficient documentation, or a platform that makes their experiments reproducible and data assets discoverable. Likewise, managers sometimes lack a way to automate reporting or facilitate data science project reviews and processes. According to O’Reilly, holistic solutions for metadata creation and management, data provenance, and data lineage are uncommon even among companies implementing AI systems.

To address the challenges, Cyril Brignone and Gregory Haardt founded Vectice, a startup that allows data science knowledge to be automatically captured and translated into metrics for operational managers and executives. Following a proof of concept, the company today announced that it raised $12.6 million, bringing its total capital raised to $15.6 million.

Capturing AI knowledge

A number of startups offer data lineage products aimed at improving governance in the enterprise. For example, Datafold automates processing workflows to maintain a baseline measure of data quality, while Solidatus offers data management and modeling tools aimed at data scientists and engineers. But Brignone and Haardt assert Vectice is differentiated by its wider scope.

Brignone, a serial entrepreneur and former research manager at Hewlett-Packard Laboratories, teamed up with Haardt to launch Vectice in 2020. Haardt was previously the CTO and VP of engineering at Lattice Engines and spent several years in product manager roles at Apigee and Salesforce.

“A handful of leading AI companies spent years developing internal solutions for their own data science knowledge and team management. Unfortunately, those solutions are custom and not available to the market. We built the same kind of solution but for all enterprises,” Brignone told VentureBeat via email. “In most organizations, AI project knowledge is locked within AI platforms. This knowledge is not accessible by the management and stakeholders and therefore not actionable. They also struggle with fragmented knowledge, almost all enterprises use multiple AI platforms and tools. [Vectice competes] with the status quo for those companies.”

Vectice’s platform plugs into existing systems and auto-captures the assets that data science teams create, including datasets, code, notebooks, models, and model training runs. It then generates documentation from business requirements to production deployments with version and lineage information, allowing users to retrieve any assets and learnings produced across multiple frameworks and libraries.

“We complement [existing project management systems] with several benefits,” Brignone said. “[Vectice] automatically captures and documents data science team knowledge on datasets, code, learnings, experiments, documentation, and projects. [It also helps to] onboard employees quickly and avoids tribal knowledge loss during project transfer or team member departure, [simplifying reviews by] defining repeatable best practices, establishing review processes, and promoting knowledge sharing. [This improves] alignment with business stakeholders by … showcasing impactful projects and successfully deployed models.”

Documenting data science

With data management remaining a major obstacle to AI expansion in the enterprise — in 2019, more than half of respondents to a Forrester report said that they simply didn’t know what their AI data needs were — tools like Vectice could help to smooth the path. As recently as 2018, only a third of companies in NewVantage Partners’ annual data analytics survey said that they’d succeeded in creating a data-driven culture.

In its survey, Forrester recommends that firms adopting AI build a pipeline of consequential AI use cases and invest in growing their AI engineering teams. “Data scientists are central to turning data into intelligent AI models. However, an oft-heard complaint from data scientists and businesses alike is failure to operationalize AI models,” the coauthors write. “That’s because implementing transformative AI use cases requires a broader team — an AI engineering team — consisting of data scientists, business analysts, developers, operations professionals, and project managers.”

“The Vectice platform solves wasted spendings in AI by directly addressing two of the most common reasons for which AI projects fail,” Brignone continued. “Vectice creates a unified view of the data science initiatives across an organization. In one place, team members can discover previous artifacts, share domain knowledge, and communicate project progress to stakeholders. Vectice enables new, collaborative behaviors that increase team productivity, centralize project visibility, and reduce the common risks associated with failed AI projects.”

San Francisco, California-based Vectice, which has backing from Crosslink Capital and Sorenson Ventures (who co-led the series A announced today) in addition to Spider Capital, Global Founders Capital, and Silicon Valley Bank, claims it piloted its platform with 19 Fortune 2000 companies. Brignone says the new funds will be put toward increasing the size of Vectice’s customer-facing and R&D teams to “respectively scale up our client onboarding capabilities and support new product integrations.”