Soepia was born in the classroom, not in a pitch

Why we exist

AI reached studying before education reacted. What we saw on campus was an entire generation "studying with AI" without gaining real depth, because no tool combined cognitive science with the practicality of artificial intelligence. Soepia fills that gap: it brings what research knows about active recall, spaced repetition, and staged focus inside a personalized scientific learning path.

How it started

The starting point was observing how students actually study. I talked to classmates frustrated by feeling they were only copy-pasting from chatbots, and to professors worried about protecting critical thinking without rejecting AI. From there came the long part: two years researching evidence-based teaching techniques (Roediger and Butler on retrieval practice, Cepeda on spaced repetition, Sweller on cognitive load, Bjork on desirable difficulty) and designing, with those professors, how each technique becomes a product feature. Every decision was validated with real tests on university students.

Our mission

Revolutionize education so anyone can learn anything in their own way, with personalized scientific method, as if they had a genius beside them. Not a slogan: this guides every product decision. The core is to understand each person's learning objective and build the shortest, sturdiest path to mastery.

Who it is for

University students who want to use AI without becoming dependent on it. Exam candidates who need to learn a lot of content fast, with real retention. Professionals learning new things alongside work. Self-directed learners on any subject. And professors who want to offer their students a platform aligned with cognitive science instead of a lazy classroom replacement.

Who founded Soepia

Soepia was founded by Matheus Augusto Mendonça, a Software Engineering student at PUC Campinas and Undergraduate Research fellow in Data Science and Artificial Intelligence. The research focuses on Machine Learning and Natural Language Processing applied to interpreting large volumes of data, building knowledge graphs, and creating interactive decision-support tools — exactly the engine behind Soepia personalized learning paths. LinkedIn profile.

• Undergraduate Research in Data Science and AI at PUC Campinas (Machine Learning, NLP, knowledge graphs).
• Co-founder of a startup through the PUC Campinas Acceleration Program (smart university shuttle management).
• AWS Academy Machine Learning Foundations (2024).
• EmbarcaTech program — embedded systems training.
• AIoT Lab Robotics Team — interactive robot for elderly care homes.
• Published author in "Tecnologias Digitais na Educação: Dos Limites às Possibilidades", Vol. 7, Editora Bagai (2024). Available here.

Team and partners

Soepia is developed in ongoing dialogue with PUC Campinas university professors and AIoT Lab partners who validate methodology, content, and experience. Engineering, design, and the scientific learning engine are built in-house. Product decisions follow the same yardstick we apply to study: staged focus, evidence before opinion, continuous adjustment based on real student learning.

Principles

Five commitments that show up in every part of Soepia.

• Scientific learning first: no decision enters the product without cognitive or pedagogical evidence.
• The student's objective drives the path: we start from "what do you want to learn", and the scientific method is built around that.
• Active learning, not passive: every session works retrieval, testing, and spaced repetition, not just reading.
• Transparency about AI: AI is a support tutor, not absolute source of truth. The student keeps critical thinking.
• Co-creation with the university: students and professors shape every release.

Science behind the method

Every Soepia decision rests on validated literature in cognitive science and educational psychology. The references below are the primary sources behind active recall, spaced repetition, and staged focus inside the product.

• Roediger & Butler (2011), Trends in Cognitive Sciences — retrieval practice outperforms passive rereading. DOI.
• Cepeda et al. (2006), Psychological Bulletin — quantitative synthesis of the spaced repetition effect. DOI.
• Sweller (1988), Cognitive Science — cognitive load theory, foundation of staged focus. DOI.
• Bjork & Bjork (2011), FABBS — desirable difficulties. Bjork Lab.
• Karpicke & Blunt (2011), Science — retrieval practice beats concept mapping. DOI.
• Mendonça (2024), Editora Bagai — chapter on usability applied to educational technologies. Available here.

Contact

For academic partnerships, pedagogical suggestions, support, or press, write to contato@soepia.com. We share method news and principles on the official Instagram @soepiaedu.