Germany-based, Kern AI provides tools to improve data quality & enrich data sets (from label automation to cleansing & monitoring). Founded in 2020, by Johannes Hoetter (CEO) & Henrik Wenck (COO), Kern AI launched in July 2022. Its offerings now consists of four products (the first two being open source): 1) “Refinery” (flagship product) (database + application logic) – combines training data & algorithms in a way that developers & data scientists can easily build NLP (data quality) automations, 2) “Bricks”(content library) – a collection of modular & standardized code snippets which can be directly integrated in refinery, 3) “Gates” – an online monitoring & inference API for data-centric models, 4) “Workflow” – the orchestration layer for NLP tasks that allows building complex workflows, which can be triggered by a variety of events. The company’s suite of products is used by AI-driven data scientists at e.g. Samsung, Barmenia, VHV Versicherungen, Evolution Time Critical (which uses Kern AI to “automatically detect the intent & sync requirements of incoming freight forwarding requests with their transport management system”), DocuSign, co:here & crowddev. “In some ways, Kern AI is similar to Zapier, but rather than following a rules-based approach, it’s built for more complex NL understanding”. It’s worth mentioning that the “data labeling” landscape is “busy” across proprietary – e.g. Snorkel AI (US) (€126m) – & open source – Argilla (ES) (€1.6m), Heartex (US) (€23m). So what’s Kern AI doing differently? According to mgmt “it’s the only open source & modular full stack” currently on the market; the platform can be used as an add-on into existing (label) platforms (think Label Studio by Heartex) or it can be used to build entire NLP applications from scratch. Pricing ranges from €149-999/ month, but there’s also a free tier for individual developers (access to which can be requested here). With the fresh funding, the approx. 10-person team plans to expand their platform’s feature-set to cover additional workflows including audio & document-based data, as well as build products for a much broader range of industry use cases. <Source: techcrunch.com, dev.to, future-of-computing.com, seedcamp.com, siliconcanals.com, finsmes.com>
