In our Second Session of the Edtech Startup Festival 2024 we hosted 3 lightning conversations around how we at Brighteye are approaching the field of AI in learning, speaking to some of the startups we’re inspired by.
Speakers:
Catalin Voss CTO and Co-founder @ Ello
Husayn Kassai CEO and Co-founder @ Quench.ai
Emilia Molimpakis CEO and co-founder @ Thymia
Rhys Spence Head of Research @ Brighteye
You will find a write-up of the session below, and the recording can be watched here:
The short write-up:
- AI in Edtech: AI is transforming education, addressing issues from student motivation to upskilling employees. Startups like Thymia, Ello, and Quench.ai are leveraging AI for mental health diagnostics, literacy improvement, and workplace training.
- Thymia: Uses gamified AI to diagnose mental health issues by analysing speech, behaviour, and expressions. Challenges include defining mental health "ground truths" and cultural differences in symptoms. Future plans include wearable device integration for deeper insights.
- Ello: AI-powered reading coach for kids, overcoming challenges in speech recognition and maintaining engagement. Unique in personalizing reading experiences, with plans to enhance its multimodal capabilities.
- Quench.ai: Offers a query-answering AI for workplace productivity and an AI coach for roleplay training. The query system has been successful, but the AI coach is still evolving. Future focus is on refining personalised coaching to stand out in a competitive market.
- All 3 startups: Face competition but are pushing forward with a commitment to innovation and personalisation in their respective fields.
The full write-up
AI seems poised to change education for the better, a fact reflected in the number of new Edtech startups implementing artificial intelligence systems in their products. AI-first startups aim to improve education outcomes using AI-centred solutions - they address a variety of issues, ranging from poor student motivation to inefficiencies in the market for skilled labour. Sometimes, they even try to reinvent education entirely, leveraging AI’s creative potential to challenge the pedagogical status quo. AI-second startups address similar issues, though their use of AI is not central to their core operation, preferring to supplement existing offerings using LLMs as opposed to developing proprietary AI.
So how can AI solutions revolutionise learning for all, from schools, to the workplace, to your home?
It can help:
Medical professionals seeking to diagnose hidden mental health issues, as well as individuals seeking to better understand their own mental health
Teachers trying to encourage reading, by gamifying the process and customising the topic to each child’s interests
Companies seeking to give their employees opportunities to upskill by providing them with a tailored AI coach
Brighteye firmly believes in this area of Edtech (see Brighteye’s article on “Education in the AI era” for more detail), which is why, in this session, Rhys Spence, our Head of Research, interviewed three founders whose startups focus on the three AI solutions outlined above. Here is an overview of the conversations which took place with each one of them.
Thymia
Their story
Thymia is a gamified test which allows consumers and medical professionals to diagnose mental health issues. Its proprietary AI records and listens to users, picking up on eye movements, micro-expressions, reaction time, typing behaviour and many other physical markers to predict mental strain, stress, burnout, exhaustion and more. Thymia was co-founded by Emilia Molimpakis and Stefano Goria. Previously, Emilia worked in academia, where she had been using language as a biomarker for cognitive function. After supporting a friend in her battle with poor mental health, navigating the poor support offered by traditional medicine used in the NHS and in private clinics, Emilia decided to use her background to find a new way of diagnosing mental health issues – this marked the starting point for Thymia. Today, comprising a team of experts and scientists, Thymia are at the forefront of AI in Edtech, and it was a pleasure to hear from Emilia throughout this session.
Developing AI can be hard…
Thymia faced a few challenges in developing their own AI model. First, finding a ground truth on which to train the model was difficult. While a cancer-spotting AI will typically be trained using pictures of cancer-infested cells, there isn’t always a consensus on how mental health disorders manifest themselves – as such, Emilia and her team had to conduct their own research into this area before they could begin training the model. Another issue was in making the product functional across all age ranges, ethnicities and backgrounds – there are cultural differences in the way stress manifests itself, for instance, and these need to be taken into account by the model. This is an ongoing fight for Thymia, but there is certainly progress being made every day.
Earning Trust in Thymia
Emilia also addressed the importance of gaining the trust of consumers and doctors when introducing AI into mental health diagnostics. She acknowledged that while there’s already widespread recognition of the need for better tools, scepticism around AI in healthcare remains. Traditional methods often fall short—mental health diagnoses can be wrong up to 70% of the time—so the need for innovation is clear. To address this, Emilia focuses on building credibility by ensuring that Thymia’s AI works alongside traditional diagnostic tools rather than replacing them outright. The inclusion of respected experts on the team also helps reassure both doctors and patients. Additionally, transparency is a key part of Thymia’s strategy; they openly explain how the AI works, its limitations, and how it complements human judgment. This approach helps break down the natural hesitations people have toward AI, allowing them to see it as a valuable tool in improving mental health assessments.
Competition and the Future
Thymia operates in a relatively unique space with no direct competitors in using AI to diagnose mental health. However, Emilia knows that maintaining transparency and trust is crucial to staying ahead in a rapidly evolving field. By clearly explaining how their AI works and its limitations, Thymia has set itself apart in mental health diagnostics, making it easier for users, clinicians, and investors to get on board. Looking to the future, Emilia plans to expand Thymia’s capabilities by integrating wearable device data. This move could introduce new competition, as wearables is a sizeable market with many players already dominating the space. But by combining this data with Thymia’s existing approach, Emilia is confident that Thymia can offer a more comprehensive and personalised tool for mental health assessments. Her focus is on hyperpersonalisation, developing adaptive models that cater to individual needs, from neurodiversity to specific learning styles, keeping Thymia ahead of the curve.
Ello
Their story
Ello is a startup dedicated to improving literacy rates by making reading more engaging to children. Ello is the name of the talking elephant which users interact with – it listens to reading, corrects mistakes, motivates, and tailors reading material to the user’s interests. The app uses its own proprietary AI for voice recognition, text generation and speaking, making it one of the pioneers of AI-first Edtech startups. We were delighted to welcome Catalin Voss, co-founder and CTO at Ello, to this session. Catalin founded Ello along with CEO Tom Sayer and CXO Dr. Elizabeth Adams. The need for the app was first observed by Adams, who noticed her daughter was struggling to learn how to read over Zoom during the pandemic. During this time, Stanford reported that 30% of 2nd and 3rd graders were behind their expected reading level! The need for new solutions to help children learn to read was clear, and the 3 founders combined their expertise in AI and psychology to develop Ello in response.
Voice-recognition Struggles
Ello’s journey hasn’t been without its challenges, with the biggest hurdle being the development of reliable speech recognition technology for young children. Existing models simply weren’t effective in capturing the nuances of children's voices, which meant the team had to develop their own system from the ground up. This involved not only building a model but also collecting the necessary data to train it - no small feat when working with diverse users.
Another significant challenge was ensuring the AI could recognise and adapt to the wide variety of accents, ethnicities, and speech patterns among children. This added a layer of complexity to the technology, requiring extensive testing and refinement. Additionally, keeping kids engaged with the learning process meant the team had to design an experience that was not only educational but also fun. To overcome this, Ello integrated playful elements like the elephant character to make learning feel less like a chore and more like a game. Balancing these technical and design challenges has been a continual process for the Ello team.
Convincing Parents
Ello has put considerable effort into convincing parents that their product is not only effective but also trustworthy. Understanding the natural concerns parents have around technology and data privacy, Ello offers an option to not connect to the internet, giving families control over how their data is shared. By prioritising transparency, Ello has made it clear that data sharing is entirely optional, with about 60% of parents opting in, showing growing trust in the platform. Additionally, Ello's child-friendly design and safety-first approach, which earned them recognition as one of the few child-safe AI models by Common Sense Media, further reassure parents that the platform is designed with their children’s best interests at heart. These measures demonstrate Ello’s commitment to balancing technological innovation with the safety and privacy concerns that matter most to families.
Competition and the Future
When it comes to competition, Ello is in a unique position. While there are other educational tools aimed at improving literacy, few offer the same level of personalisation and engagement that Ello provides with its AI-powered reading coach. Most competitors in the space focus heavily on assessment, whereas Ello’s strength lies in making the reading experience enjoyable and motivating for young learners. Catalin and his team are particularly excited about the future, with plans to further enhance the platform’s multimodal capabilities, allowing it to better understand and support each child’s individual learning style. Their long-term vision is to inspire a love for reading in every child, using AI not just as a tool for learning, but as a partner that adapts to and grows with the child’s progress, making literacy development more accessible, fun, and effective.
Their story
Husayn Kassai founded Quench.ai, a tailored AI coach which works with companies to give their employees opportunities to upskill. Quench.ai allows organisations to train their own tailored AI on company data, creating a virtual assistant which can help employees with any objectives they’re trying to achieve by providing detailed plans to achieve them. Husayn has previously founded a unicorn startup called Onfido, which provides ID verification services. In their early days, they had a close connection with employees, fostering engagement and development. However, as the team grew, this personal connection was lost, and productivity began to decline. This experience led to the creation of Quench.ai’s main product, a query-answering AI system designed to quickly deliver the information employees need, no matter the scale of the company. We are delighted to have Husayn join us for this session!
Varying Success across Products
In the meeting, Husayn Kassai explained the two key products offered by Quench.ai. The first is their AI-powered query answering system, which helps employees quickly find the information they need, improving productivity across large organisations. The second is their AI coach, designed to provide personalised, real-time coaching through roleplay, helping users prepare for real-life situations like meetings or sales calls. While the query answering system has been a success, widely adopted and praised for its efficiency, the AI coach hasn’t seen the same level of uptake. Despite its innovative roleplay feature, which allows employees to practice in a low-pressure setting, it has struggled to really take off. Still, Quench.ai continues to refine the product, and hopes that the product can soon live up to its potential.
Competition and the Future
Quench.ai operates in a competitive space, especially with major players like Microsoft and OpenAI also focusing on AI-driven workplace solutions. While these tech giants have deep resources and a strong presence, Quench.ai sets itself apart through its focus on personalisation and real-time coaching. However, the competition is stiff, particularly as larger companies refine their AI tools to better cater to user needs. Husayn acknowledged this challenge but emphasised that Quench.ai’s commitment to hyperpersonalisation gives it an edge in creating more tailored learning and productivity experiences.
Looking ahead, Quench.ai has ambitious plans. The company is focused on evolving its AI coach to better address the broader challenges that have been harder to tackle, like improving communication and general skill development. There’s also potential for expanding its query answering system, which has already been well received, into new industries. With the backing of strong engagement from EdTech firms, Quench.ai is set on refining its products to stay ahead of the curve.
And with that, we’ve come to the end of this session! We greatly appreciated this discussion with Husayn, Catalin and Emilia.
Comentários