Sophya

Future of work skills are becoming increasingly important, and technological ‘whiplash’ very often disproportionality affects the disempowered. With the current education infrastructure, millions of learners are left behind, without equitable access to basic education or critical upskilling - unable to keep up with education/skill demand, despite education being key to inclusive economic growth.

The internet holds the answer to scalable learning, but is too unstructured for learners who need guidance in learning the most. Internet content will continue to become more open, free, and germane to skills of tomorrow, so, informed by data science analysis of internet learners, we're building the 'Spotify' of learning, where learners no longer have to go find the right content - the right content will come and find them. We are specifically working with a celebrated Bangladeshi professor, Dr. Naveed Alam, to ensure culturally-appropriate equitable access to the best possible online education throughout metro, regional, and rural Bangladesh.

According to the World Economic Forum’s latest report on Future of Jobs, by the mid-2020s, 75 million jobs may be displaced by technology trends, while 133 million new roles may emerge. Separately, as the world’s population continues to increase, brick and mortar educational institutions cannot keep up with global demand for basic education or these new technological skills. 

Both massive problems above disproportionality affects underserved populations. This is partially because underserved communities fill many of the jobs that will be replaced, and partially because these same communities suffer structural disadvantages in pursuing the basic education and further training that the new roles require. 

We are tackling the problem of inaccessible free education and approachable retraining that would be beneficial for underserved and middle-class parts of Bangladeshi society, particularly given its' rapidly increasing internet penetration (~40% now), and its' young workforce.

Consistent with global learner trends, we found that aspiring and current students were often confused about how to reach their professional goals. (“I want to be a programmer/doctor, but don’t know how”). We solve these problems by combining the 'power of the crowd' with data science, to construct always-improving Learning Paths for popular subjects that people can freely use to learn from.

We ultimately aim to serve learners from ‘pre-K to gray’, addressing the needs of the entire education and workforce sector. But initially, we’re starting in health science and computer programming for students who simply have an internet connection. We chose these starting points because skills in health and programming are sought-after, ‘future-proof’, and are key to economic prosperity in burgeoning societies. There is also a shortage of millions of healthcare and technical workers that will worsen indefinitely without a scalable system for learning these skills.

We’re actively working with students in these fields to crowdsource the best free and/or open learning content on the web that is actually meaningfully helpful to them in career or learning progression. I.e. the internet content they actually use to learn. We’re also working with experts and faculty in these fields to see what is recommended from the ‘top down’, to cross reference with the material that students are organically finding, sharing, and using on their own.

Using this data, we construct recommendation systems that give similar learners exactly what they need, in the right order that they need it. Students report far less wasted time, confusing career pathways, and barriers to learning desired skills.

We describe what we’re building as ‘the Spotify of Learning’, because instead of having to go out on the web and sink hours into finding the right content, we built a system where the right content instead comes to the learner. At the right level, and in the right order to learn best. This is informed by data from how community-members are learning.

Ultimately, Sophya enables anyone, anywhere, to skill-up to become anything they want, for free, using the power of the Internet. Any person can input their level of education  (e.g. Grade 4), and where they want to go (e.g. ‘be a doctor/engineer/data scientist/roboticist’). Like a GPS for learning, Sophya displays an interactive pathway to take, with each step being a learning node containing the content needed to progress. This content is crowdsourced from the web, comprising the most effective, up-to-date resources that similar learners organically use. This enables learners to flexibly move through learning, on or offline, from any device, without having to be actively in attendance at school, anywhere in the world. Because content is crowdsourced, Sophya will be able to return location- and language-specific content that is most likely to be helpful for a given learner in their specific niche. When people in given communities learn in Sophya, it provides a network effect to local learners: serving up content likely relevant given their proximity to other learners nearby. This is partially why we are working with renowned surgeon-educator Dr. Naveed Alam from Bangladesh.

Sophya also provides AI tools that enhance learning for students. For one, we enable learners to curate and organize their learning content into solo or collaborative Learning Paths (where we also use machine learning algorithms to recommend helpful ‘missing’ content). We provide tools to enable learners to interact with their content to enhance learning. For example, our computer vision and natural language note-taking tools allow learners to write notes or draw directly onto video they are watching (live!), and to export their results. They’re thus able to create strong memory anchors and learn better. 

Part of improving learning is improving student engagement. We provisionally patented our highly engaging 'foreshadowing' technology, that alerts learners of content that will occur in the _future_ of a video, so when they write notes onto the video, they know where they should or shouldn't write. 

New processes:

• Application of data-driven crowdsourced recommendation for learning. 

  • We built a recommendation system that aggregates content from students and teachers in the content area it is relevant to (e.g. recommended health content is derived specifically from health students/teacher).

  • Statistical models applied to algorithms, weighting content appropriately. This ensures that expert-curated content is more likely to be shown to learners, but despite this, content that students frequently use/find helpful is also shown.

  • Pathways displayed to Learners update in real-time, based on use by target learner group. This ensures that learners have access to the most useful content.

New technologies:

• We created (and have new patents pending on) several novel tools that allow learners to engage deeply and interactively with existing content on the web.

• Includes our video note-taking tools (which allow students to write their notes directly onto live video and export the results), our video-object detection tools, our flashcard tools, and our automated assessment tools.

Existing technologies and processes:

• Algorithms that identify and group users into clusters based on content they engage with (i.e. Spotify-like, but applied to learning).

• Pinterest-like boards where content is aggregated and displayed.

  NB: Why does this all matter? For lowering barriers to learning. Briefly: i) Learning paths/boards with recommendations take fatigue/confusion out of finding the right stuff to get you from point A to B. ii) Tools to increase engagement and learning result in more stickiness/pleasure for the user, which drives learning longevity.

From our current pilots (>15,000 students), students in and outside of school are giving us feedback such as the following:

“This saves me SO MUCH TIME! I can finally study and learn the way I actually want to.” - Non-school learner.

“I save so much time using the app. And the timestamped notes and automatic PDF give me exactly what I need for review with no work. THANK YOU!” - Female STEM school learner.

“TIME SAVER! Seriously this provides a more efficient way to take notes, especially with the auto-pause feature I can pick up right where I left off.” - Male premedical student.

Theory of Change: 

1. People in any given community use Sophya to have a data-validated, peer-curated stream of education/upskilling resources provided to them.

2. Users jump onto Sophya, input their current and desired levels of education or upskilling. 

3. Sophya provides an interactive pathway of resources from the web, in the right order needed to learn the desired skills.

4. Peoples’ use of resources within the platform helps Sophya to know whether or not to keep that material in the system for subsequent learners. 

5. Because of peers in the community who use Sophya to learn similar skills, new learners get resources more closely matched to their local needs.

6. Learners don’t have to be in school or pay tuition. Therefore, underserved communities gain access to upskilling immediately.

7. Learn/upskill → better work opportunities, or can choose entrepreneurship to create wealth and jobs → change socioeconomic status → reinforce positive cycle for community.

Our general principle is to build software that encourages collaborative, peer-enhanced learning. This is particularly important because there is no way that educational institutions are going to be able to keep up with the demand by students for basic or advanced future of work skills, and if peers do not have a platform in which to reliably learn together scalably, billions of students/student hopefuls globally will inevitably be left behind. Furthermore, encouraging peer-collaboration helps scale to more users at lower cost to the company, which aids in sustainability. 

Currently, we serve approximately 5,000 out-of-school students, and 10,000 in-school students. We have close to 100,000 students in our pipeline to join the platform (via relationships with institutions). This lines up well with our growth projections to have 100,000 students using the platform within 1 year, and close to 4 million students within 5 years. 

We generally divide our goals up into a few buckets: overall/mission, growth, technical, partnerships.

Next year:

• Overall: Plant the seed in the minds of communities without easy access to good schools that free, efficient learning on the web can be done with the right system in place (Sophya).
• Growth: We’re aiming to have 100,000 students/lifelong learners on the platform within the next year.
• Technical: Within the next year we aim to have full-scale recommendation systems for healthcare and technical ‘Future of Work’ content (e.g. robotics, programming, data science).
• Partnerships: We’re aiming to partner with at least 5 leading public institutions who can help endorse content and progression for open access Learning Paths.

Five years:

• Overall: Be the go-to platform for people around the world to use whenever they want to learn anything in the most efficient way. Give everyone on the planet the ability to make maximal use of their ‘brain capital’.
• Growth: We’re aiming to have close to 4 million students/lifelong learners on the platform within 5 years. 
• Technical: Within five years, we aim to have recommendation systems for K-12 education, several undergraduate programs (including all available STEM education pathways), and much more robust healthcare and ‘Future of Work’ learning pathways.
• Partnerships: We’re aiming to partner with at least 50 leading public institutions who can help endorse content and progression for open access Learning Paths.

Next year:

• Model training: We want to ensure our models for recommendation are neither over or underfitted, but serve the initial beachhead market of future of work training well. 
• Distribution: Once product-market fit is reached, our key task will be distribution.
• Legal: We have to ensure that content being embedded or uploaded isn’t posing any copyright issues. We can follow in YouTube/Pinterest’s footsteps (both multi-billion dollar companies, so they clearly have Terms of Services that the law is okay with).
• Financial: Have to hit our milestones without running out of cash.

Five years:

• Model training: continually improving, but also new skills will become available 
• Distribution: Similar as before - but instead of distributing through institutions, we’re able to deliver directly to students who ideally share with each other.
• Market: The market will evolve as competitors enter and required upskilling changes.
• Culture: As we expand geographically - we’ll have to contend with places that don’t see education for women of color or particular class systems as important as for the ‘dominant’ group. We need to break through this.

Each can be difficult, but is doable. We’ll lean on our resources, which include a diverse and accomplished advisor board, our partners, data, government and non-government organizations, and our ‘smart creative’ teammates.

• Model training: We’ll ensure we have a diverse set of users and narrow datasets on which to train our learner-level-specific algorithms. We’ll ensure enough learners are on the platform so that the algorithms are solid.
• Distribution: We have MOUs with multiple schools with student numbers that total in the hundreds of thousands. We also have social media growth hacks to help with distribution. We will make the software as shareable and fun as possible to encourage distribution by our users.
• Legal: Why, lawyers of course! Get creative with building high value to the user without IP issues.
• Financial: Raising enough now to get to sustainability and then profitability. Then we have the option of being revenue growth or can raise a venture round to scale as fast as needed. 
• Market: Ensure we keep an ear to the ground re: market dynamics, and new future of work skills needed.
• Culture: Still working this out. But from now we’re building relationships with two major NGOs that could help us enter culturally different markets in a sensitive way.

Our plan is for initial expansion into Bangladesh in the next 6-8 months. We aim to build a recommendation system for the Bangladeshi context by working in partnership with key Universities in Bangladesh to first train the Sophya algorithms on content proven to be helpful to actual Bangladeshi students. The first University that we aim to work with is the University of Dhaka. We then aim to use the recommendation algorithms pre-trained in this way, combined with 'best of the web' learning content in each given field to then distribute to other Universities and schools in regional Bangladesh. This way, students will get a blend of content proven to be helpful in a local context, combined with content useful by similar learners worldwide.

Bangladesh is an incredible market opportunity for engaging with students, upskillers, and retrainers in our Future of Work era. According to the WEF, close to 50% of the population of Bangladesh will live in towns and cities, >110m people are already connected to fast internet, and the country is being rapidly urbanized. Over 60% of the population is young, and will need a scalable system of learning that can flex to their context. We have this system!

We have 7 full-time employees (CEO/COO and five engineers), 3 part time data scientists, and 3 interns (UX design, market research, and growth). 

We have experience in education, finance, growth, programming, data science, research, public health, and medicine. These competencies are strongly aligned with building a sound, scalable system of education. We are also advised by Dr. Naveed Alam, a renowned Bangladeshi surgeon-educator working with us to bring Sophya to Bangladesh.

Our leadership team is composed of a medical doctor (Vishal) and a PhD student (Emma) at Harvard. Each were popular Teaching Fellows at Khan Academy, and have worked on building education/data platforms at the World Health Organization and at Harvard Medical School. Vishal (CEO) helped scale a prior startup’s userbase from tens of thousands to hundreds of thousands of students. Emma (COO) is an insanely talented co-founder who knows how to product manage, fundraise, manage the team, hire, and code. Our CTO Mark has 20 years of software engineering experience and leads the engineering team with care, humor, and great mentorship.

We care about the learner’s experience with the software. When people can personalize, collaborate, or just enjoy interacting with an app, they spend more time with it. We’re user-centered-design focused, and weight user experience on par with desired outcome. We have an in-house design team, front-end specialist, and UX designer. Our short iteration cycles center around testing with students.

We’re surrounded by an incredible team of advisors/investors from Harvard, MIT, Google, Stanford, Carnegie Mellon, and Facebook, who span several relevant verticals including education, artificial intelligence, data science, and business operations. We just graduated from Harvard’s top incubator (Launch Lab X).

Sophya is currently partnering with 4 large universities across 2 countries for algorithm training and initial piloting. Three of these universities are using Sophya primarily in their health sciences schools, and two are preparing to deploy Sophya university-wide in the coming months. The pilots are focused around improving recommendation algorithms for students, and around user experience testing, feedback, and iteration.

In the spirit of improving health education access, health science faculties are helping to train the ML recommendations on appropriate content that should be distributed to all health science learners - both in school and out. The university-wide deployment is aimed at ‘not overfitting’ - ensuring that the tools and platform are suitable for many types of learners. 

The total number of students who will be able to access the software from these pilots is approximately 100,000.

We’re SaaS. We provide some of our services to students/learners for a monthly or annual subscription fee, and some of our services for free.

Because of its potential importance in creating sustainable education and skills equity, we provide our learning recommendation technology for free, worldwide. But, students can opt to upgrade to premium accounts, which cost $5-7/month/student. This allows them to have unlimited use of our previously described AI tools to enhance their learning experience and to save time, and allows them to upload learning content for personal use, and to share content that they have permission to share.

Institutions can pay us to provide insight into their students/employees skills analyses, skill gaps, and assessment metrics.

While we’ve currently raised enough investment capital to sustain us for 1 year from today at our current burn rate, we aim to be a revenue driven company, and should become cashflow positive from user revenue by January 2020.

To be clear, our current model is fee-for-service, and our customers are students (subscribing to our enhanced learning tools), and institutions (paying for analytics and insight). In the future, we may explore an additional market-linkage model, allowing students to flag topics they need help with and having tutors join the platform from anywhere in the world, to provide help for a fee.

The more we can understand what free and open content on the web students are using to learn in international contexts, the more we can do a good job of constructing hot-updating learning paths and making career pathways visible to underserved communities, at scale. While we’re tackling this problem by partnering with institutions around the world, we would be able to build the best models primarily by accelerating partnerships with institutions. Tiger could help in the following ways:

i) Assistance with partnerships: This piece is CRITICAL and working with Tiger would be incredibly helpful. As described above, users of Sophya benefit from a strong network effect. And, our algorithms rank most highly the educational content that is recommended by local experts (e.g. teachers in a given space or students who are doing well). Therefore, Tiger could be incredibly helpful by helping us build relationships with educational institutions in South Asia whose experts could help validate the internet content that their local students would be using to learn. We then can much more quickly train algorithms to provide location-helpful educational pathways to any learner.

ii) Community/network: This is also a very valuable piece. Communities of early-stage entrepreneurs are incredibly powerful because we each deal with similar issues, and can help each other get through them.

iii) Funding: Money is any startup’s oxygen. The grant and access to other funding is great.

iv) Personalized support: Mentorship, a brain trust of advisors, and specific help with PR would all be very helpful.

In general, we’d like to partner with educational institutions (at any level) to help validate and train models to be helpful for local learners who may or may not be in school.

We would particularly like to partner with Universities in Bangladesh, so we can do a really good job of ensuring the learning content we recommend is suitable for its context. In addition, we would be keen to partner with educational institutions in underserved areas. That way we can do two things: i) ensure that local community-members who aren’t able to enroll in those schools can get vetted educational internet content germane to their region, and ii) ensure that we can cross-reference the internet material that is being validated at schools in underserved areas with schools in high economic status areas in Bangladesh to look for significant differences in material. We can then recommend content in a more equitable way while keeping it relevant to the social context.