Amsterdam-based KIMO, an AI-powered intelligent learning platform, has announced closing a €3M funding round. The round led by two Dutch investors comes thirteen months after an initial round of €1.25M closed in September last year. With the funding, KIMO will further scale its personalised learning platform for tech domains such as AI, Blockchain and Cloud Computing.
The Dutch AI firm is just 2.5 years old and was founded by two Harvard graduates who met in Boston during their executive MBA programme. As lifelong learners, they realised the need to change the way current programmes work to reskill the world’s population. Their idea culminated in KIMO, a platform offering personalised and continuous learning at scale.
“In KIMO, we believe the game-changer will be personalised education – made possible by the AI revolution – which aims to serve the user exactly the right content at the right time,” said Rens ter Weijde, CEO of KIMO.
KIMO plans to use the additional capital to grow the user base, strengthen the team, and set up the right partnerships. The firm recently announced that Vijay Mehra, a former Leader of McKinsey Digital Labs in Asia, will be its new CTO. Kent Haverstock, formerly a Digital Strategy Activation Lead at Bose, will join as Head of Growth. Currently, it’s actively looking for AI talent.
How does KIMO differ from other Edtech platforms
While content aggregators such as Coursera and Udemy rely on bringing content from many educators onto a single platform, KIMO differs by including other content formats. It is a modern platform that works like a combination of YouTube, podcasting app like Pocket Casts, or News aggregators like Apple News. The difference is immediately visible when you sign up for the service.
Once you create your account, KIMO shows the six domains it supports right now. This includes AI, Blockchain, Cloud Computing, Cybersecurity, Computer Vision and Digital Marketing. Once you select the domain, KIMO organises the content into topics and sub-topics. These contents are presented in the form of an overview which makes it easier to decide whether you want to learn a particular topic.
The content layer is further diversified with formats such as news articles, videos, medium posts and podcasts. KIMO is essentially the equivalent of bringing content related to a common domain from a wide range of sources. This is, of course, different from traditional platforms that model their learning system similar to a university degree and experience a similar dropout rate.
In an exclusive conversation with Ai.nl, Rens ter Weijde said that “the system is set up to be modular as one of the key advantages to more traditionalist education approaches”. This approach allows KIMO to build new domains faster. “It takes less than a week in principle to build a new domain (e.g. crawling, clustering, indexing, moving towards recommendations). Quantum computing, Data Science, Fintech, Robotics, and possibly the ‘Metaverse’ are also coming in the next few months,” he added. With its domains, KIMO is focusing on the digital realm and upskilling people in a critical knowledge area.
User-centric: How AI changes KIMO’s learning platform
Rens explains that KIMO is designed to be a user-centric platform instead of content-centric. Since the challenge of building a continuous and evolving learning platform is immense, CTO Krishna Nallamilli made universal scalability another prerequisite for KIMO. They added a third layer to it in the form of habitualisation. The founders of KIMO believe that effective learning should be a habitual thing and not something you do once.
With the strategy combining user-centric, universal scalability and habitual focus defined, KIMO built its AI platform. The first part of this exercise is understanding a knowledge domain. The engineers at KIMO rely on transformer models and graph representations to understand any domain. This is in stark contrast to the traditional model, which relies heavily on experts and is subjective. This makes it difficult to scale and not offer precise recommendations.
Rens and Krishna also saw the current landscape where the skills needed in a workforce are constantly evolving. So, they relied on transformer models which can change rapidly and built a model that uses bottom-up scalability by learning from the web. Once the domain is built, KIMO uses the latest techniques in the AI field to “build highly detailed knowledge graphs that capture the structure of knowledge domains in great detail.”
The second part of its AI platform is to link the domain available on its platform with a specific job. The idea here is that skills are more valuable and people with the right skills will become employable. So, they used millions of job postings on the web to recognise nearly 50,000 skills and were further enhanced by linking them with automation risk and salaries offered in different regions. This skills recognition per job can be described as the killer feature of KIMO.
In a nutshell, KIMO is relying on an AI model that is powered by machine learning and one that relies on data. While KIMO’s platform is ready, Rens told ai.nl that they still consider it to be beta. He confirmed that there are more than 110,000 users and KIMO considers them beta testers since the founders feel the need to “improve in the product by making the service faster and reliable on a wide number of devices and improving the quality of recommendations.”
“The target is [to reach] 2 to 3 million users by next year and have an online learning data, cross-platform, at scale,” he added.
Is curation the cure for online learning fatigue
KIMO has a platform capable of being user-centric and is constantly evolving thanks to the AI model. It aims to trump other EdTech platforms with its curation but Rens feels that they are “not there yet with the curation”. It offers two streams of curation – expert-led in the form of playlists or webinars and AI-led. The AI-led curation relies on a vector matching system combining contents and users vectorised on a number of dimensions.
With the vector matching system, the labels such as clicks, like and watch time are fed into the system along with other dimensions such as the popularity of content online and age. “My feeling is we will need nearly 1 million people using the app for a while in our critical segments to make sure this works to our satisfaction,” Rens told ai.nl. “We are playing with more AI models to change the way people learn, such as real-time QA models (eg replacing the teacher, trained on all the underlying data we sit on, free), instant summaries (visible in the beta app), and jobs-to-skills inference to allow tailored content for jobs,” he added.
A partnership with India’s Reliance Jio
Last month, Reliance Jio, an Indian telecommunications company, started selling its first smartphone under the JioPhone brand. Called JioPhone Next, the budget smartphone is aimed at first-time smartphone users and those looking to switch from a feature phone. While the JioPhone Next is expensive compared to similarly priced phones from Xiaomi, Infinix, and others, it wants to be all about curated experience.
One such experience will be KIMO’s continuous learning platform. KIMO CEO Rens ter Weijde confirmed that the company and its shareholders see Jio as a perfect partner to bring high-quality education at scale to emerging markets such as India. “We’re discussing launching KIMO together on the new JioPhone Next,” he said without sharing details on availability. KIMO is also working with an Indian college, Google and Huawei to bring its platform to emerging markets.
Edtech is the next big frontier for RIL and the company is building a suite of diversified learning solutions for India’s edtech market. It acquired a 73 per cent stake in Embibe in 2018 with the goal of becoming top player in India’s test preparation segment. It is also said to be building AI-backed services. With KIMO, Reliance might be able to bring a service that has already built a reliable AI platform for online, continuous learning.