EduBeyond

EduBeyond is solving the issue of educational disparity on a global level. Given recent efforts in increasing the accessibility of education, the percentage of world adolescents out of school has decreased to 17%. However, education disparity remains a prominent issue even for those with access to basic education. 

In Indonesia, where we have been operating for the past year, the public education infrastructure has been one that is "enforced" but not necessarily effective. This is evident in only 16% of young Indonesian adults being able to transition into tertiary education despite a 98% and 79% enrollment rate at the primary and junior secondary levels, respectively. 

With adolescents making up around 28% of Indonesia's population, it is essential to evaluate the topic of education beyond the accessibility baseline. This is shown by the stagnant tertiary education rate lagging behind the increasing enrollment rate. As public funding for education increases, adequate infrastructure is essential to ensure a significant impact on welfare.

Education effectiveness/quality, or lack thereof, can be attributed to factors such as lack of proper learning materials, low-quality teacher training, and inequitable resources due to varying income classes. These factors, and many more, have been exacerbated by the recent pandemic, which was the catalyst for an abrupt change in educational content delivery. 

To put the pandemic's disruption of learning into perspective, a study by Mckinsey showed that the graduating class of 2021 lost, on average, five months of teaching materials, with an accumulated total of 2 trillion teaching hours lost worldwide. Even during digital instructional hours, "stunted behavioural and socioemotional development" and "chronic absenteeism from students" are common problems in many education systems. 

These statistics indicate that while there are sufficient digital infrastructures for classes to be conducted during the pandemic, the lack of personalization and engagement factors has created adverse consequences for the effectiveness and efficiency of content delivery. 

Equitable, quality education allows for an increased high-skill workforce, a significant increase in lifetime earnings, and added bridges for globalization. As Indonesia has the fourth largest education system in the world, an effective intervention tool would result in a larger talent influx to advance the world economy and innovative solutions for global crises.

With regional governments increasing their educational spending on physical infrastructures, the advancement of technology and curriculums must catch up to create quantifiable improvements in welfare. Given the estimated 1.5 billion primary and secondary learners worldwide, effective digital interventions must be implemented systematically to prevent a loss of potentially $10 trillion in lifetime earnings. 

Uninformed pedagogical methods, untested curriculums, and inadequate digital infrastructures are some of the major factors why the tertiary education rate is notably lower in underserved communities. Further, students from our target population come from a long line of poverty, resulting in a lack of incentives to pursue higher education even if the opportunity is present due to financial pressure. 

Thus we have developed an interface that can be incorporated with any curriculum at any grade level between K and 12. With the foundation of a standard digital classroom, we have integrated generative AI applications to make course administration effective and efficient. 

Our digital interface is similar to that of Microsoft Teams and Canvas, including basic features such as classroom chats, WebRTC, assignment upload, and discussion forums. 

Homework and practice problems are calibrated to each student's progression and proficiency with the subject, we have developed an adaptive feedback system with variants of the Leitner method. Depending on the student's "feedback" with each practice question, the algorithm can adjust the retention intervals for optimization. 

To cut down on time spent on tasks such as test creation, marking, and course administration, we have created an auto-marking system (optical character recognition) and a test creation tool (natural language processing) to allow teachers to spend more time on personal interactions with students. 

Additionally, EduBeyond provides a chatbot that provides academic support during non-instruction hours. Currently, the chatbot can give hints and recommendations based on student inputs. 

Our technical solution aims to accelerate the academic progression of low-income students while lowering their required instructional hours. Using generative AI features such as our chatbot and question recommendation systems, we hope to retain more students past the compulsory education duration while offering them better preparation for high-skill jobs. 

With the support of product research backed by the University of Toronto, quantifiable improvement in learning results will be shared publicly. In order to make curriculums more effective in the acquisition of civic knowledge and community engagement, our curriculums are developed under the guidance of education professionals from partner universities and local schooling communities. 

Since 2020, we have been working with students on the outskirts of Jakarta and Surabaya. While having access to basic education, these students are behind in their English and technological proficiency by almost one academic year compared to their counterparts in private and higher-end public schools. Given the recent pandemic’s disruption of course administration, the gap has widened significantly.

As mentioned in the “specific problem” prompt, low-quality teachers, untested curriculums, and inadequate digital infrastructures are some of the major factors why the tertiary education rate is notably lower in rural regions. Further, many of the students we currently serve come from a long line of life in poverty, resulting in a lack of incentives to pursue higher education even if the opportunity is present. 

To put it in quantitative terms, the average school day in Indonesia is 10 hours, with a mandatory 12 years of compulsory education. Compared to the United States, where the school day is 4 hours shorter and almost 50 ranks higher in quality (according to the World Population Review). With this in mind, more than 90% of low-income students would elect directly into low-paying labour jobs in hopes of providing financial assistance to their families. 

The lack of quality education goes beyond socio-economic consequences, students we have interacted with in the past often feel a disconnect with relevant issues within the community, lacking the necessary social-emotional and relevant information to make a difference. 

Through the analysis of qualitative research based on student demographics, data can be used to improve the relevancy of implemented curriculums and technical features.  

The team’s co-founders are Tien-lan Sun, Miklos Sunario, and Alec Shi. 

Tien-lan is EduBeyond’s CTO and a freshman at UC Berkeley’s MET program. Having won first place at the International Science and Engineering Fair (ISEF) in the Translational Medicine category with his novel AI fundus camera system, Tien-lan is leading the development of our interface, with a particular focus on the adaptive question algorithm and the AI-powered chatbot. Tien-lan is working with Berkeley’s ML@B in training EduBeyond’s ML model. 

Miklos Sunario is EduBeyond’s COO and a second-year at the University of British Columbia. Having experienced the Indonesian education system, Miklos is a key player in our consultation with local students and teachers in Indonesia. He is responsible for many of our Indonesia-based progressions, such as an invitation to present our product to the Indonesian Parliament and having our computer science curriculum be certified by Indonesia’s Ministry of education. 

Alec Shi is EduBeyond’s CEO and a second-year student at the University of Toronto. Alec is responsible for EduBeyond’s internal governance and external applications. One of the applications led to EduBeyond being selected as an award winner by the Moonshot Platform, a social impact foundation powered by the Avast Foundation. As a result, the founding team was able to present their project to the Deputy Secretary General of the UN, Amina Jane Mohammed. Alec has also been invited to a legal incubator hosted by McCarthy Tétrault, the leading Canadian law firm in international trade practices. 

We are currently deploying our interface within local charities and home-schooling communities in Indonesia. With decisions pending from schools in Indonesia for monetary contracts, we believe that our solution is more of a prototype than a concept. 

EduBeyond can really benefit from an increase in beta testers, both students and educational professionals. With an increased beta-testing population, qualitative analysis can be more effective, given the diversity of participants. 

Cultural liaisons are also extremely helpful in the context of curriculum building and providing relevant examples within educational content (for example analysis of a culturally relevant story to teach grammar)

EduBeyond's solution has two notable differentiating factors in comparison to other ed tech companies in the field: 

- Application of academic research in product development

- Catered to students of low-income classes and developing regions

EduBeyond's interface features will be evaluated via random controlled trials with our user base. With support from the education faculty at the University of Toronto and Berkeley's Machine Learning Lab, we will identify the effectiveness of integrated features based on pre-established impact metrics and user feedback. 

Currently, the major players within the Southeast Asian edtech industry have prioritized their focus on retaining students instead of producing quantifiable improvements in learning; EduBeyond hopes to break this trend by exhibiting transparency through our research results.   

The results from our product research will be made public to advance innovations in the digital education space.

We are also the first edtech company to prioritize distribution in lower-income classes. This is done through our B2G model instead of charging the product per individual. The B2G model relieves any financial burden students may incur from subscription-based ed tech products.

Our company identity emphasizes social impact, which led to our current negotiations with the Indonesian government to distribute our product as a third-party market vendor. Since major players within the Southeast Asian edtech industry cater to the private sector via individual subscriptions, the public sector remains a relatively unoccupied space.

• Have 5,000 active users between May 1st and August 1st

  • Beta testing phase 1

  • With private partners (NGOs, charities, private schools)

  • An estimate of 3 months (end date could change slightly)

  • Establish weekly feedback meetings with students and teachers

  • Bi-weekly interface maintenance

    • Immediate intervention in case of emergency

• Have 10,000 active users between September 15th and March 15th

  • Beta testing phase 2

  • With iterations from Phase 1

  • Private partners and homeschooling students 

  • Six months duration (start and end dates are estimates)

  • Automated feedback system + feedback meetings with teachers

  • Weekly interface maintenance 

    • Immediate intervention in case of emergency 

Research (1 year)

• Conduct A/B testing with focus groups (100 students each) for the following features (research structure in development)

  • Adaptive question recommendation algorithm 

  • Qualitative analysis of interactions with chatbot

• Timeline

  • Trial 1: Beta testing phase 1 (3 months)

  • Product iteration phase (1 month) 

  • Trial 2: Beta testing phase 2 (6 months)

Research (5 years)

• Published academic paper on tested features with the support of

  • Education faculty at the University of Toronto

  • Berkeley ML@B

• Continue A/B testing with additional AI-based learning tools (if deployed)

Reach

• 20,000 students in 2023-2024

  • 10,000 users from private partners we are currently working with

  • An estimated 10,000 from initial deployment through the Indonesian government

    • Focusing on a select number of school districts as a pilot

• By 2028-2029

  • 30 million students reached (estimate, reasoning below)

    • 56 million youths between 12-24 in Indonesia

    • With around 40 million in public education infrastructures 

Education metrics for our cohort (in 5 years)

• Tertiary education rate of 70% 

  • Continued product R&D similar to the beta testing schedule

  • Periodic testing with standardized materials 

• Graduation rate of 90

  • Around 54% of low-income classes as of 2022

  • Track time spent on interface in relation to academic scores

    • Personalized intervention, if necessary

  • Periodic feedback from teachers on

    • Curriculum

    • Interface and generative AI features

Quantitative: 

• Tertiary education rate
• Graduation rate 
• Number of instructional hours vs non-instructional hours spent by teachers
• Standardized testing scores

Qualitative:

• Basic demographics (gender, income, parent education, etc)
• Post graduation employment
• Conversations with chatbot and instructors
• Surveyed responses to product features and curriculum content

Activities: 

• Technology cuts down administrative task time for teachers
  • More time for personal teaching and guidance
• Better quality and cultural adaptation of curriculum 
  • Support from U of Toronto, Berkeley, and local education professionals

Outputs:

• Teachers now have more time to focus on the progress of individual students
• Students get a more personalized education experience 
  • Technology and increased teacher attention

Short-Term Outcomes:

• Students gain better civic awareness and general knowledge
  • Easier to become change-makers within the local community
  • Wide range of job options
• Teacher's time is allocated more efficiently without worries about employment risks

Long-Term Outcomes:

• The upward trend of a community with improved
  • Critical thinking skills
  • Social-emotional awareness
  • Technical knowledge relevant to community issues
  • Socio-economic status
• A new model of course administration 
  • Use technology (with human inputs) to enhance the level of engagement and content relevancy instead of diminishing it

Most of EduBeyond's AI learning tools were created using the Keras and OpenCV libraries and OpenAI's GPT-3 application programming interface. 

Instant communication between users in a chat system was implemented with Socket.IO, the Express.js application programming interface (hosting the socket), and MongoDB (storing messages). 

The class module used the same technologies but included openGraphScraper for embeds. The improved Leitner system used MongoDB aggregations with TypeScript and Express, js. 

NextAuth.js was used for authentication. AWS's S3 was used for file storage, and AWS Lambda was used for deployment. 

The video conferencing system used WebRTC (communication protocol), Janus (handling WebRTC connections and acting as a selective forwarding unit), and WebSockets (for breakout room communication). 

Our employees stem from 4 different countries and 8 different cities. Recruitment typically happens at various social impact conferences (United States, Sweden, Czech Republic, etc), which allows for a diverse pool of applicants. 

For EduBeyond, we believe that team diversity is the most authentic when it occurs naturally, and our team demographics are largely attributed to our international presence. 

EduBeyond is pursuing a B2G and B2B model. We will act as a third-party market vendor and distribute our interface to schools and school districts, with the latter done through the Indonesian government. 

This model is ideal in a few ways. For one, EduBeyond can generate sufficient revenue for its incurred expenses through government funding while ensuring that students won't have to pay individually for the product, unlike most ed tech products that rely on subscriptions. 

Additionally, our current arrangement with the Indonesian government offers us a degree of autonomy to prioritize school districts for distribution. This allows us to target infrastructures with relatively lower graduation and tertiary education rates, making a more significant difference in the general welfare. 

From a commercial standpoint, our official customers will be the Indonesian government and private schools, which will pay for our product in components. In addition to the education software, we are also building an introductory English curriculum, which would be licensed separately. 

While the contractual agreements are still under negotiations, a five-year term would be ideal as it presents a sufficient time frame to consult with our user base and conduct meaningful research on user experience. 

Before deployment, EduBeyond will be sustained through grants, philanthropists, and private partners. 

Once deployed, we will be engaged as a third-party market vendor by licensing our product with the Indonesian government and private schools. Indonesia's Ministry of education and our private school partners will support the deployment budget. We hope to adopt a similar model with countries like Malaysia and Thailand. 

We are also pursuing funding domestically from Canada's International Trade Department, specifically an infrastructure grant to export Canadian technology abroad. 

EduBeyond's path to financial sustainability is integrated within its goal for social impact, and it will be able to self-sustain its operations to accomplish the mentioned impact goals.

Moonshot Platform

• Gave grant award of $10,000
• Referral to speak at the UN ECOSOC forum and present to DSG Mohammed

Delta Fellowship

• $10,000 award

Atlas Fellowship

• $10,000 award

Penabur Schools

• Private school chain in Indonesia
• Under negotiation for three year term contract