75% increase in subscriber conversion rate with Gemini through API
95% reduction in AI task failure rate
3x faster response times
67% reduction in AI costs with Google Cloud
Algor Education switched to Gemini models to build a fast, personalized AI learning platform, boosting conversions by 75%.
Algor Education switched to Gemini models to build a fast, personalized AI learning platform, boosting conversions by 75%.
Improving platform speed and accuracy with Gemini
We chose Gemini models because the balance of latency, cost, and performance was one of the best in the market. The high rate limits mean we are always covered during peak hours, reducing fallbacks in our backend and allowing us to design more complex study tools.
Emanuele Gusso
Co-founder and Head of AI, Algor Education
For many students, the format of a lesson can be just as challenging as the subject matter itself. Traditional educational materials, such as dense textbooks, don’t suit all learning approaches. For students who think visually or live with dyslexia, reading large blocks of text can make studying difficult. While personalized, one-to-one tutoring solves this by tailoring the approach to the student, it’s too expensive to be accessible for everyone. Algor Education is changing this by using AI to adapt study materials to the individual needs and learning approaches of every user.
Previously, Algor Education used a third-party AI provider, but the system's unreliability and slow response times created friction. With certain tasks taking up to 18 seconds to generate a response, students would at times assume the platform wasn’t working properly, and abandon their study sessions.
The previous AI models were also prone to hallucination. In the field of education, where accurate information is essential, this unreliable output risked damaging user trust in the platform. At the same time, Algor Education’s previous provider struggled with multiple LLM calls during busy exam periods, meaning the engineering team lacked the confidence to build new, advanced features.
Seeking better performance and accuracy to solve these issues, the company transitioned its AI workloads to the Gemini API. As a participant in the Google for Startups Cloud program, Algor Education benefitted from the support of the Google Cloud team to experiment with and build complex new AI pipelines on Google Cloud. The faster response times of Gemini models enabled Algor Education to keep students engaged, while the accuracy of its output meant the platform could handle complex personalization requests without hallucinations.
“We chose Gemini models because the balance of latency, cost, and performance was one of the best in the market,” explains Emanuele Gusso, co-founder and head of AI at Algor Education. “The high rate limits mean we are always covered during peak hours, reducing fallbacks in our backend and allowing us to design more complex study tools.”
Transforming textbooks into personalized study tools with Gemini
Gemini models adhere perfectly to the strict structural schemas we require for concept maps, while at the same time offering total flexibility on the content. Combined with the specific voices of the Text-to-Speech AI, students can now listen to complex study materials in a highly tailored format.
Emanuele Gusso
Co-founder and Head of AI, Algor Education
Algor Education allows students and teachers to upload their study materials, such as textbooks, video lectures, and audio recordings. The system then processes these files to provide customized educational resources, including interactive concept maps, summaries, and flashcards, helping users absorb the information in a format that works for them.
To make this happen, the platform relies on Gemini through API. First, the system standardizes the uploaded textbooks and media into a common knowledge base. The platform uses Gemini Embedding to mathematically calculate which passages of text are related, while Gemini Flash-Lite reasons out the specific contextual relationships between them.
Once the material is mapped, the platform uses Gemini models to construct strict hierarchical structures—like a conceptual tree for a mind map—based on the user's specific learning needs. Algor Education allows learners and teachers to personalize the material to suit their requirements, such as asking the AI to simplify a specific chapter or explain a complex topic using metaphors. The Algor Education team uses prompt engineering to ensure the models remain firmly anchored to the original uploaded texts, maintaining accuracy while adapting the style to suit the learner.
The platform also offers an audio-based study option, using Text-to-Speech AI to turn the generated text into structured educational podcasts. The Algor Education team used the tool to select specific voices and direct the tone of the output for an optimal learning experience for its users. Finally, an evaluation phase uses Gemini to generate self-assessment quizzes and conversational chat tools, allowing students to test their knowledge interactively.
“Gemini models adhere perfectly to the strict structural schemas we require for concept maps, while at the same time offering total flexibility on the content,” says Gusso. “Combined with the specific voices of Text-to-Speech AI, students can now listen to complex study materials in a highly tailored format.”
Scaling accessible education and improving application speed with Google Cloud
By migrating to Google Cloud, Algor Education significantly improved the performance of its platform. The AI task failure rate dropped from a peak of 2% to below 0.1%, giving the engineering team the confidence to develop complex features without worrying about system reliability. The median wait time for a task also decreased from 18 seconds to six seconds, helping to reduce user frustration and bounce rates. At the same time, Algor Education saw a 67% reduction in their overall AI costs. This faster, more reliable user experience has helped increase the rate of users converting from free trials to paid subscriptions by 75%.
Algor Education is now focusing on expanding into new markets across Europe. At the same time, the company is developing new features, including the capability to connect students with private tutors.
Personalized, one-to-one education has always felt like a utopia because it’s expensive and difficult to scale. Google Cloud has allowed us to make it a reality, providing an accessible, highly personalized learning assistant for students and teachers everywhere.
Emanuele Gusso
Co-founder and Head of AI, Algor Education
It’s also exploring the use of Google Cloud speech-to-speech models to enable vocal interactions during the student evaluation phase, further broadening how students can interact with their study materials.
“Personalized, one-to-one education has always felt like a utopia because it’s expensive and difficult to scale,” concludes Gusso. “Google Cloud has allowed us to make it a reality, providing an accessible, highly personalized learning assistant for students and teachers everywhere.”
Algor Education is an educational platform that uses AI to transform unstructured study materials into interactive, visual, and personalized learning tools that adapt to how individual students learn.
Industry: Education
Location: Italy
Products: Google Cloud, Gemini, Gemini Embedding, Gemini Flash-Lite, Text-to-Speech AI
