Founded in 2021 by Anubhav Mishra and Aprit Jain, ZuAI raised $484K from Prime Venture Partners in its seed round last week
The startup is deploying its take on OpenAI’s GPT to build an interactive experience for students and is currently developing a model for maths problems and other equations
The generative AI opportunity is undoubtedly massive and early stage edtech startups such as ZuAI have their task cut out as even larger players are eyeing this space
According to a 2021 UNESCO report, India had a huge disparity in terms of the pupil-teacher ratio (PTR) in 2018-19, which stood at 26:1, i.e., just one teacher for every 26 students.
This ratio was even worse for higher classes, secondary and senior secondary schools, at 43:1 and 47:1, respectively, a far cry from the 35:1 specified in the Right to Education Act.
Can AI improve this situation? That’s what Bengaluru-based edtech startup ZuAI is betting on.
Anubhav Mishra and Arpit Jain, alumni of VIT and IIM Bangalore’s Launchpad programme, founded ZuAI in 2021 to leverage AI models to solve the tutor problem.
“We see generative AI to be one of the pivotal technologies to solve this particular problem, both in terms of accessibility, affordability and also the quality of content, which is being generated,” Mishra said about ZuAI’s mission.
The startup started building its products with the simple aim of bridging the gap between providing an affordable, quality education and simultaneously giving a one-on-one tutoring experience. Last week, ZuAI raised $484K (INR 4 Cr) in a seed round led by the early stage investor Prime Venture Partners, and the founders believe that the GPT opportunity will change the game in online learning.
Having onboarded nearly 4,500 daily active users across classes IV to XII, the startup is looking to build on OpenAI’s large language models to build an interactive experience for students, calling it an ‘AI tutor’.
ZuAI — GPT With An Edtech Twist
Talking about how ZuAI works under the hood, Jain said the company has deployed the Generative Pre-trained Transformer (GPT) technology developed by US-based tech giant OpenAI to start things off.
The large language model (LLM), however, has certain ‘restrictions’ such as it can only generate textual content for Indian students from classes IV to XII across all streams and subjects.
According to Jain, the platform uses 3-4 language models, and depending on the requests, one or more of them are active at the same time, enabling a high degree of personalisation.
ZuAI’s app offers three chatbots – Exam Bot, Revision Bot and Class Bot. The Exam Bot quizzes the student on topics of their choice, the Revision Bot helps with explanations of key concepts, and the Class Bot allows students to take notes from specific chapters.
“On the back end, we decide which particular model is fit for the problems that each student faces and based on that we use a particular model for that student,” Jain said.
The startup has also baked difficulty levels from 1 to 10 into the platform, where a higher number is a higher difficulty. These levels are automatically triggered based on the responses received from a student. The founders explained that the model starts on a medium difficulty, and increases as the student gets better at a subject.
Students can also force the AI into asking questions of higher difficulty levels, but ZuAI plans to restrict such behaviour as far as possible to retain guided learning to some extent.
There is one problem that ZuAI’s GPT-centric platform cannot solve. It cannot generate content for all the subjects yet as mathematical solutions are not the strong suit of AI models currently. ChatGPT and other LLMs suffer from glitches or hallucinations where they output wrong answers.
Currently, the Bengaluru-based startup is working to develop an in-house computational model to work on numerical-based problems for physics, chemistry and mathematics lessons. So far, though, ZuAI’s focus is completely on creating text-based lessons that explain concepts in these subjects that require calculations.
“We are not seeing that solution [mathematics] because we have already restricted it to curriculum content, to only answer and respond from the curriculum content and not beyond that. In mathematics, we are working on something, on building our own LLP model to solve the thing step by step,” explained Jain.
Despite raising its first round of funding, ZuAI has not yet solved the monetisation problem. The product is currently free to use, but this could change in the future. The edtech plans to introduce a subscription model for users once it has garnered a sizable user base.
Will AI Become Edtech’s Playbook?
ZuAI is far from the first edtech company to leverage ChatGPT and other LLMs to build course content or track outcomes.
Even the likes of BYJU’S and Unacademy are launching products with a generative AI flavour and it is easy to predict that more and more startups will take advantage of gen AI in the future.
For instance, BYJU’S has developed BADRI, an in-house AI model, for personalised learning. Additionally, BYJU’S WIZ, a suite of three AI transformer models, leverages BADRI and two other models — MathGPT and TeacherGPT — for a greater degree of personalisation and delivering effective outcomes.
This will be the challenge for companies such as ZuAI, which have taken the first steps into the wide generative AI pool. Larger players can quickly erode any early mover advantage due to their scale and deeper pockets.
One of the ways that ZuAI and other early stage edtech startups can get a leg up on the bigger competition is by differentiating with India-focused context features. This could mean integrating with government programmes for early and higher education or using AI for skill development and other areas in edtech.
Early Days For Generative AI
While investors have naturally grabbed the generative AI opportunity, the reality is that new models and new LLMs require significant hardware resources for development.
AI-generated video startup Hippo Video cofounder and CEO Karthi Mariappan believes that the hardware costs alone are very high for most startups because of how many discrete graphics processors are needed to generate new models.
Developing a global generative AI success story from India means dealing with the ground reality of the high capital needs for AI startups, according to All In Capital founder Kushal Bhagia.
In an earlier conversation with Inc42, ZuAI backer Prime Venture Partners’ managing partner Sanjay Swamy said that AI can change the game for incumbents in people-reliant sectors such as fintech, consumer services and edtech and replace human resource-intensive models that are proving hard to scale up sustainably.
He also indicated that gen AI could change the whole unit economics game for startups, particularly in terms of needing fewer human resources to pull off a bigger scale. We all know that edtech giants are suffering from a slew of issues ranging from slow revenue growth to high customer acquisition costs and an uncertain transition to hybrid models.
Indeed, this is the age where more and more edtech giants are looking to supplement online learning with physical centres. Can ZuAI’s AI-centric model and its focus on online-only learning actually become a competitive moat?
With inputs from Nikhil Subramaniam