Turma1

Turma1 uses Learning Management Systems and Artificial Intelligence to create a individual experience in a class of students.

Usually, the precious time of a teacher is spent lecturing the maximum possible amount of contents, without any feedback on how it was absorbed. On another hand, the student has little time to practice what is learned and acquire the necessary skills to be proficient in a subject.

 The burden of real involvement with the content is left in charge of the student when alone, favoring unfocused attention and procrastination. Furthermore, those in struggling stay for a long time unnoticed, until a test is applied.

 The frequency of exam applications is low, once they take too much class time and are labor intensive in elaboration and correction. Exams explore too many subjects through a low number of questions covering only a small subset of the content, besides be executed in a moment far away from the lecture.  

This approach has two main negative effects:  

1. Many students study only for the tests, usually on the eve, being the only moment in which they interact with the content.

2. After poor grades, the long time interval between evaluations directly impacts the daily motivation to persevere in the content, shaking their self-confidence. Furthermore, once that knowledge will not be evaluated anymore even if is a required for other subjects, the gap will persist.  

Turma1 believes that using Blended Artificial Intelligence Learning (BAIL), a mix of Blended Learning, Artificial Intelligence and use of Learning Management Systems (LMS), there are no more reasons to be attached to the traditional learning paradigm.  

In the Turma1 proposal, between 70% to 75% of time spent during class is applying the content through problem-solving. The usual day is divided into four quarters: 1-lecture, 2-with peers and teacher, 3-alone, 4-teacher solving for the class. Daily, our students solve between 30 to 50 problems, almost a 1000/month.  

To feed questions for each day, we apply Deep Learning algorithms to automatically identify question statements representative of a subject in a massive database of public exams and contests. Through LMS, we registry each student answer and track his performance. Beside this, we estimate the performance of the student on related questions not answered and compare it with data available on university entrance exams. Everyday is test day!  

Teacher’s time in this approach has a more noble function: make connections with the student and motivate him. The immediate and continuous feedback makes it easy to identify which students need more attention and in which aspect of the content. The progressive adaptation of the difficulty of the questions presented  become a motivator for students, since it behaves like a game, generating a sense of accomplishment at the end of the day. The opportunity to help each other and the section of teaching solving fix their understanding.

We believe that the success of our students is a positive impact of our work in the society and could drive attention to our approach, creating a blueprint for other institutions to follow, helping the whole sector of education in our community, our country and globally.

LMS are very useful in delivery content and registering user interaction, but feeding them with content and classify each content in segmented domains for a more effective evaluation of which skills should be reinforced is labor intensive.

During 3 years of graduate studies, we researched ways to do this task in a more effective way. Our solution uses recent developments in Deep Learning as Deep Knowledge Tracing (Piech et al, 2015) and a improved version of a text classifier that we published in IJCNN 2018 - A Compact Encoding for Efficient Character-level Deep Text Classification.

Our vision as an education startup is register and evaluate every interaction that our students have with the LMS to identify learner difficulties in early stages of his work, motivate them through constant feedback and give them a daily sense of goal achievement. Our idea is to concentrate teacher effort on the most common mistakes presented by the class and spend more time with students struggling with the content, leaving more advanced learners to explore the content. Technology is also a great ally to analyze results available in public data to  compare our student's results with a better perspective.

We are working to integrate our solution to the LMS in a more seamless approach, lowering the necessary expertise to operate the whole system.

At the same time, we are working hard with our students to obtaining good results in university admission exams and scientific competitions, looking to drive attention to our approach, get funding to a bigger structure and more marketing, allowing us to grow our number of students served.

That way, we could finance more research and development of our approach, consolidating the process in a product and create partnerships with others schools.

Yearly in Brazil, 9 million students apply to 200k vacancies in university admission exams. Courses that prepare students for these exams are very traditional business, and a high number of students approved is a powerful driver of growth, since it is understood as a leading of quality. We are betting that with the power of modern technologies we could be more efficient and have a real chance to compete. We plan to use an unit as a showcase to expand our methodology to many other schools through partnerships, providing funds to expand our research, produce content and developing software.

Our strategy is grow our number of students marketing our results in social networks platforms and buying spaces in local outdoors. Our objective is help the maximum amount of students to achieve approval in admission exams and use the credibility created to expand our methodology to others unities and partners schools. This is not an unusual strategy, since others courses already used it, but we think that using technology, we could have more control and standardize the quality of process between model showcase and school partners, better than just providing booklets and educational contents, as others they did.

We started our pilot in mid 2017, using our own savings to buy tablets, furniture and pay rent. Then, we could afford to serve only 6 teenagers and 3 kids with problems in math. From them, 4 were approved in federal universities and 2 kids were finalists of the OBMEP (Brazilian Mathematics Olympiad).

This year we are serving 10 teenagers and 8 kids. Based on the data that we daily collect, we are very confident of theirs performances in admission exams and results in OBMEP. We consider our pilot was successful and ready to grow into a bigger number of students.

In 2019 we expect to serve 30 students, the limit of our infrastructure today. We are looking for fund capital to structure an unit that afford 150/210 students by year, distributed in classes of 30. We operate in a city with 500k habitants, and there are many others courses and schools with the same amount of students, so we think that is a very realist number. Our intent is invest heavily to make that unit a success showcase to spread across the our metropolitan area (12MM habitants) and later whole the country (200 MM), through partnerships or franchising.

Our team started with a PhD dropout in Communication and an MSc in Computer Science that are passionate by teaching. Our vision is to be recognized as an edutech company of excellence, providing affordable software solutions and educational contents that help teachers extract better results with his students. By being research-oriented and dealing with a so important sector as education, we think that we could attract many like-minded individuals: scientists to research area and the best educators to produce contents. Through the marketing of our results, we expect to create partnerships with others schools or franchising to selected owners.

Since we only have a small pilot classroom, the source of revenue that we received until now were a small tuition payments from our students ($150) that allowed us to pay rent, bills, buy tablets and study our methodology in practice. We already had good results last year, and expect impressive results this year too.

Our objective is drive attention to our approach and raise capital to fund a bigger showcase unit. Traditionally, education in Brazil has serious quality problems, so will not be difficult to positioning ourselves as a elite school, that usually have high tuition prices. But we think that are more to be done.

We plan to use the revenue of this showcase unit to  invest massively on technology and production of educational contents, spreading good practices across many units. That way we could offer to many students better materials and leave to the local teacher the more noble task of helping the ones that are in struggle and motivate grow.

Using the same platform technologies, we could identify units that need extra support and redirect our human and capital resources..

We really believe that we could impact the sector and offer a better alternative in education.

We think that just by being chosen as a Solve Team, it will drive attention to our approach and establish our brand as a serious player in the market, easing access to capital to finance our expansion in better terms. The grant alone could  allow us afford more students, do more research, develop more technologies and spread our work.

With our savings and the tuition of ours students we could not afford an expansion on the velocity that we desire. With the projection of Solve, we expect reinforce our credibility to raise funding and accelerate this process.

We think that with a bigger operation, we could grow our pilot into a showcase unit and transform our methodology in a formal high school, with more stable font of revenue for many years, helping us to finance our expansion. There is demand for good high schools in our area, especially with a technological appeal as we meant to be.