AI powered English lessons for Indian classrooms

India's schools cannot offer good quality English lessons because they can’t find good teachers. According to a report by UNICEF, more than 50% of children in South Asia live in learning poverty, which means they are unable to read and understand a simple text by age 10 (https://www.unicef.org/rosa/what-we-do/education). Those who can speak fluent English go to higher paid teaching jobs at international schools,  or to jobs outside education. This problem is especially acute outside of big cities - there are few good English teachers in tier 2 locations (https://www.unicef.org/rosa/children-south-asia). . 

In India, only 40% of the students who attend these schools can access online resources and less than half of them can afford it, and self driven app learning - which is normally designed for wealthy western adult learners -  is ineffective for Asian primary school aged students. 

The result is that students fail to learn English, and schools and local tutors fail to build great local Education businesses, despite the large underlying demand for learning English. 

Our product changes the game for both. We are the first product built for low income schools in Asia - we are low tech, our product works with novice teachers and students, works offline, and at low cost. 

Our software allows teachers who are novices (pedagogically, and language skills wise) to teach international quality English in their local school. The software delivers an international quality English lesson using voice to text software, and the teacher can pause the lesson for correction and focus students on topics as suits them. The software also shows students how to correctly speak words like a native speaker. 

The students - who cannot afford more expensive teachers or tech solutions - get to access an international quality English programme in their local school, for prices that their parents can afford.

We have built a teacher app and a student app that we sell to low cost schools and tutors and NGOs to give every child in South and SouthEast Asia the chance to learn to speak English. 

It cots just $5 per student per year, and the only device needed in the entire school is the teachers' basic smartphone (these are nearly ubiquitous for teachers now). Students do not need any devices. 

The teacher app delivers research driven, adaptive Direct Instruction lessons with international accent and pronunciation. It has accompanying images (to allow for learning by dual coding) and can be operated by a teacher using just a basic mobile phone. 

This works by the app delivering scripted content, that is adapted (instantly) by the app using machine learning, to present the students with content suited to their level of learning and response to other content. It is powered by our speech to text engine, the only one built for and trained on young Indian voices (other, western trained, speech to text software cannot understand our students). 

We looking to complement, not replace teachers - at any stage the teacher can pause, or redirect, the lesson using their phone. 

The student app is an AI driven English speaking practice tool. Children can go home to a family that does not speak any English, take their parents' low cost smartphone, and have a conversation in English in which they get feedback on their vocab, pronunciation and grammar. The speaking practice is linked to what they learned in class (where their teacher used the teaching app). 

We offer the teacher and student app to low cost schools, NGOs and low cost tutoring centres in India, Afghanistan, Pakistan and Vietnam. The primary focus currently is in India, particularly Maharashtra, Telangana and Andhra Pradesh. 

Our solution serves low income communities in India, in particular children in school and their teachers. It will impact their lives by transforming their job and education opportunities by allowing them to access those that require English (the majority and best - currently out of their reach, creating inter generational cycles of poverty). 

Children in school - become fluent English speakers thanks to receiving AI powered, adaptive learning in English in the school they can afford and access. 

Teachers - their job is made easier and less time consuming by deploying our AI powered technology, which they control and direct. *Unlike other innovations, we are looking to complement teachers, not replace them*. 

The important context  is that the majority of children in south and southeast asia attend very low cost private schooling or tutoring. We believe that these schools and tutors are the best way to reach them - NOT via self driven learning. 

Self driven learning - repeatedly fails in practice and is shown to be ineffective in theory (teacher led learning, not self driven learning, is much more attuned to how children learn, especially when they are novices). 

Our theory of change is that if underserved Indian and global low income children receive *both* high quality classroom instruction *and* high quality out of class practice -  and those tools are built using their perspective and demand, and improved based on continuous data collection and analysis  - the outcome will be that they speak English well enough as adults to get better jobs, education opportunities and enjoy enhanced social mobility. The impact will be their breaking the “English opportunity gap” enforced cycle of intergenerational poverty for them and their families. 

The inputs and activities are that children in underserved Indian communities receive daily evidence driven Direct Instruction lessons in their local school, via a simple to use android app used in class by their teacher. They engage in daily English conversation practice, via an AI driven app on their parents phone, which is the first that can understand their accent and use adaptive learning (powered by constant data collection and input of their perspective) to have a conversation with them that builds their fluency. 

Our senior team are both from, and their professional career has made them highly proximate to, the communities that we will be serving. We directly engage communities in the design, development and implementation of our solution. 

Our team, and its proximity to our communities 

Sandeep Mallareddy, Founder, attended an affordable school in  India, before becoming a teacher in a low income school in Pune as part of one of the inaugural Teach For India cohorts. Sandeep has spent the last 15 years working with low income schools, teachers and students in India, as part of his work for several Indian not for profits and social enterprises. 

Ravi Singaram attended an affordable private school in India, before embarking on a software development and management career focused on building for lower and middle income Indians. He has been working closely with Indian students and teachers for three years as a result of his work on English Quest. 

Jitesh Kumar attended an affordable school in India, before becoming a teacher in a low income school in Hyderabad through Teach for India. Jitesh has spent the next ten years working with low income communities across India, as part of his work for several not for profits committed to increasing equity and access in Indian education. 

Jamie Martin has spent the last five years on the ground in India working with students and teachers in low income areas of Hyderabad, Mumbai, Bangalore and Chennai, as well as NGO sponsored students in rural areas India wide. He previously worked with low income communities across sub saharan Africa (as founder of www.injini.africa), and across the UK as a Special Adviser to the UK Secretary of State for Education. 

How we engage users in product design, development and implementation

Our student and teacher apps are the first to be built based on the perspective and need of under served south asian students and teachers. Our team has taught and worked in these communities for over a decade - some of our team are from these communities. Before building our first version we spoke to school leaders, researchers and other experts in and members of these communities. We hold regular focus groups with both teacher and student users. We have a free comment feedback box on the app for them to input feedback. Our technical team road tests all new features with students and teachers in schools. 

Equity 

We are dedicated to closing the English opportunity gap across South Asia, then globally. We have built, and will continue to build, spoken English software at a price point, on technology, and with a user experience that is based on the needs and input of under served South Asian students. Our team regularly visits schools, and calls students, teachers and parents, to solicit their feedback. We do this in an open, non hierarchical way travelling to their spaces where they feel comfortable, and where we can truly understand their context.

We have built our teacher application, and an accompanying student practice application, and we have tested over 18 months in schools across South Asia. 

Currently, our applications are used in 141 schools, NGOs and tutoring organisations in India, Afghanistan, Pakistan and Vietnam, supporting over 1,000 teachers and nearly 60,000 students. 

The school year in South Asia begins in June, at the moment we are on course to more than double our scale for school year 2024/25 - we will work with over 125,000 students, over 2,500 teachers, in over 250 schools, NGOs and tutoring centres.

We are joining Solve for the a) peer network, for the b) technical expertise you can connect us with, and for c) support with government and NGO partnerships to make impact at scale d) Measurement, evidence gathering and insight sharing 

A. Peer network 

We are excited to benefit from and contribute to a peer network of fearless innovators focused on the world's most challenging problems in its most challenging contexts. We often find that commonplace innovation or technical guidance does not suit the challenging context in which we work and our users live. We are excited to help, and in turn be helped by, a community of innovators who are grappling with the same problems as us. 

B. Technical expertise 

We are looking to use large language models and AI in a way that is cutting edge in two respects: 

1. Using AI/ML engineering to teach an *adaptive* lesson that - depending on student responses - takes the lesson in a different direction. 

2. Doing this with young Indian (and other non western) voices. Most voice recognition AI is trained on western voices, so we are being highly innovative and differentiated - but also building one of the world's most important data training sets - in developing this trained on young Indian voices. 

This will require cutting edge technical expertise, and we recognise that solve can connect us to some of the world's foremost AI & ML experts. We would be honoured and excited to take their advice in pushing the technology's frontiers to help the world's poorest children learn. 

C. Scale support, especially with governments and NGOs 

We are passionate about scaling our impact as much as we can. We believe that millions, perhaps 1 billion, children world wide can build a better family for themselves, their family and their community if they can speak English. 

While school by school (and school chain) acquisition can reach millions of students one day, to truly reach scale and truly reach the very poorest students (whose families cannot afford even our modest costs) we will need to partner with large NGOs and governments. 

The incredible Solver network, and influence and brand of MIT, can help us do this. Connections for trials or meetings with large NGOs and government officials interested in Solve's innovations can get us impact scale at a speed and size impossible on our own. 

D. Measurement, evidence gathering and insight sharing 

We are committed to advancing scientific understanding of how AI can be used to enhance learning, particularly in the field of language learning, and particularly in under served and understudied non western countries. 

We would love to work with Solve, and use MIT's expertise, to conduct experiments and studies (eg an RCT and quasi RCT) of our solution. 

We would be committed to open sharing all the data and findings, to advance the knowledge of other innovators and the use of technology to help students around the world. 

How your solution approaches the problem in a new or significantly improved way

We are the only product utilising artificial intelligence to change the learning experience of low income students *in the classroom* in India, and the only ones focusing it on spoken English. 

Our solution allows low income students to access interactive, effective teaching in spoken English (because our AI powered app delivers the content, and adapts the lesson to student responses). The teacher is able to allow the AI app to dictate the lesson, or take control and either teach herself or get the app to focus on a certain lesson path (eg repeating an exercise, or switching to teach something the students are struggling with). 

Others are applying AI in wealthier markets (less impact), or focusing it on student driven learning (less effective, as evidence shows) or on subjects which have less depth and breadth of impact than spoken English. 

We are the only product trained on young Asian students' voices (nearly all other language AIs are being trained on western voices), with a user experience and price their teachers and schools can access.

How could it catalyze broader positive impacts from others in this space?

If we can develop AI technology that can compute non western voices accurately , and can then move adaptively through content to teach students what they most need, then this can be applied to other subjects. We could support the teaching of Maths, or science, or other languages using our AI software breakthrough. 

It could also encourage others to do the same in a variety of countries and context and a variety of languages (not just schools, not just English!). This could mean a spurt in AI classroom language innovation globally, so that more people globally from all backgrounds can speak to each other. 

Furthermore, If students' listening and speaking in English improves, it will open up the developing world's students to more education content on a wider variety of problems (over 60% of online learning content is in English). Creating bigger audiences, it will encourage more innovators to enter the space and create online learning content for the developing world. 

 How could it change the market/landscape?

Currently the innovation in the EdTech market is focused on rich, individual learners  - because that is where the money is. 

Lower income students and their families do not have buying power, and often do not have devices, so no one is building cutting edge EdTech content for them. 

By building for schools and classrooms, where there is buying power owing to the number of students - and by proving that innovation at very low cost can be successful - we will change the market into one where more innovators enter to build for schools in low income settings, thus getting cutting edge EdTech innovation to low income learners. 

India is the world's biggest school market - us building for it can help to transform many others ! 

Our product will break the inter-generational cycle of poverty driven by not speaking English, which affects hundreds of millions of lives in India and across the developing world. 

Learning to speak English will change the life path of not just our students, but their families. As fluent English speakers, our students will have better education and employment opportunities. They will earn more, enjoy more satisfying careers and live in cleaner safer places - and their (extended) family will enjoy these benefits as well. 

THEIR children - the next generation - will go to better schools and have English spoken at home, creating a virtuous circle of education and employment opportunity to replace inter-generational poverty.  

This impact does not even rely on perfect implementation or outlier results - even partial English performance will have a huge impact. 

At every level of South and South East Asian society, improving your English leads to higher wages(22-34% more), better working conditions, and more opportunities. 

We believe that there is also a national and global impact of our product. We believe that the next generation containing substantially more English speakers will mean a richer, more democratic, safer South and SouthEast Asia. 

Given the geographical, demographic and economic importance of that region - an English speaking South and South East Asia means a safer, richer, more open world. 

You can read our theory of change in detail here https://drive.google.com/file/...

The impact we wish to have is to get under served learners in India and globally to speak fluent English. This is so that they can access better education and job opportunities, and break inter generational poverty and equity cycles for them and their family. 

This will track SDG goals 1) no poverty , 4 Quality Education, 5 gender equality and 8 Decent work and economic growth. 

Our specific learning target is to get students taught using our applications to progress in speaking English at the target pace of a Cambridge Assessment International English course, measured on the CEFR scale (C2 proficient in 1,200 learning hours).

We have developed an assessment schedule to show if students are on course for this. At each low stakes formative assessment (taken every 15 hours, or 90 times across their learning journey) , we assess whether a student is mastering enough material to be on course to be C2 proficient in 1,200 hours. 

We only consider a class to be successful if 80% of students are hitting this benchmark (while also staging data based interventions with students, teachers , of the other 20%). 

Across over 100,000 assessments in 2023, over 80% of English Quest students where achieving this benchmark. 

We use artificial intelligence and machine learning to develop a mobile application for teachers, and students. 

The teacher application uses AI and machine learning to compliment teachers in delivering an interactive, adaptive international quality English lesson to their class. Teachers can direct, and pause, the app at any time - it complements them, but they control the lesson. 

The student application allows students to undertake interactive practice at home, giving them live feedback on their spoken English. 

We use Open AI’s Whisper and its wider machine learning capability to create a speech to text engine that is a step change on current models and to create an English speaking teaching, practice and assessment designed for the needs of underserved South Asian learners. 

We're committed to sharing our learnings and anonymized data with the OpenAI community. This will contribute to improving Whisper's understanding of non-native English speakers and benefit education-focused AI tools in the future.

We will prioritise data privacy and security.  We'll implement anonymization techniques and adhere to ethical data handling practices

On the specific tools we’ll build:

Speech to text engine: we will Fine-tune Whisper with young Indian English voices for accurate transcription & pronunciation feedback. This will improve its accuracy in transcribing common Indian accent features, leading to better learning experiences for students.

English classroom teaching tool: We're developing a module to handle group voice feedback from classrooms. Identify participation patterns & attribute transcripts for holistic feedback and correction.

English practice: Analyse speech patterns to recommend relevant content/resources based on individual needs. We will help Open AI’s technology do vocabulary enrichment, by training it on educational materials to enhance understanding of learning conversations & exercises. It will be trained on our in house content based on researched learning patterns. We will also train it to analyse speech patterns to recommend relevant content/resources based on individual needs.

English assessments: Integrate pronunciation assessment & fluency analysis to personalise learning for students.

Across all areas, we'll implement an MLOps pipeline to integrate new data continuously from our app usage. This data will include various accents, vocabulary used in specific lessons, and user feedback, improving the STT's performance over time.

We have 19 full time team members. 

They break down as: 

3x senior leaders (CEO, CTO, COO) 

6x software engineers 

6x product managers (support schools in implementation) 

3x sales 

1 learning outcomes/engineering

We have been working on this for four years. 

We began running pilots and developing the idea in May 2020. 

We incorporated, and began working on it full time in December 2020. 

We have equality, diversity and inclusion at the heart of our mission. It is a focus during delivery, governance, in terms of our team and advisory board, and in our work with stakeholders and in engaging end users.

Our core aim is increasing equality by improving low income people's ability to learn English. We will build a product that aids inclusion for English second language learners, by making them as users a core part of our production creation and iteration. Our diverse team (and diversity friendly hiring practices) will ensure a range of perspectives and experiences are used to create a product that's inclusive for a wide range of end users - from Indian schools and students, to refugees and migrants in the US, EU or UK from all nations and backgrounds, to lower income English learners  globally.

Our team leadership is 80% minority ethnic and 30% female. In terms of new hires, we will ensure to actively seek applicants who are from backgrounds not traditionally included in artificial intelligence work, and maintain equality of access to working with us. We prioritise remote working, which should help those with caring responsibilities, or a disability, to apply successfully and succeed with us. Our lead technical advisor (also an investor) on this project is from a minority ethnic background.

In terms of our governance,  our senior advisor Alok Gupta is from a minority ethnic background. The board of our company is 50% non white, and 75% English as a second language speakers (an important non-included group in this sector). Our recruitment and stakeholder engagement policies not only spell out a commitment to equality, diversity and inclusion - we take active measures to eliminate bias from our hiring (for example, including blind scored assessment tasks where the candidate's identity is not known).

Stakeholder and end user engagement is an area of major contribution to equality, diversity and inclusion for our project. Our product will be used by students from lower income backgrounds globally, who are learning to speak English. Our user interfaces and product design will be empathetic to them (works offline, very simple user experience, highly pictorial, empathetic to disabled users). We already work with refugee agencies in Afghanistan, and will focus on engagement with refugee stakeholders in particular organisations that can help get our product to refugees as end users and aid their inclusion.

Our business model 

Our main business model is to sell our teacher application to low cost schools (~95% of revenue). Normally, the schools themselves pay for our application. Sometimes they raise the fees (so parents indirectly pay), sometimes schools pay for it our of discretionary budgets. We have some NGOs who sponsor schools, or tutoring organisations, to use the product (~5% of revenue).

We typically charge on a per student basis. We charge Rs500 ($6) per student per year. A typical school buys 400 student licences, spending 2 lakhs rupees ($2,400) with us per year. 

What products or services do you provide them? 

We give schools access to our student app for all of their teachers, and to our student app for all of their students. We give them access to our assessments that run 4x per year. 

The student app is only used at home on parents' mobile phones. The school does not require any hardware to use our product (beyond the teachers' personal mobile phone). 

This gives them a completely scripted, interactive and adaptive English lesson for every class, and interactive and adaptive at home practice, for every student. We also provide assessments that support and track student progress (4x per year). 

We provide both initial and ongoing training in how to use the product (please see below). 

We provide data based tracking and reporting of student progress - monthly reports to principals, quarterly reports to parents. We use these to intervene with classes, teachers or students who are falling behind. 

We provide support in putting on an impact showcase at the school, where students show parents what they have learned in English. This is acruc

How do you provide these products or services? 

We give schools and students log ins for our app and can give access virtually. 

Our team visits the school to train the teachers (6 hours of initial training), and then depending on school location, we engage in continuous coaching (for remote schools online monthly, for city centre schools in person monthly). 

The reports on teacher performance and student performance are automated and sent to principals monthly.

Why do they want or need them?

Schools want, and face a parental expectation, to teach children to speak English, but they struggle to afford high quality english programmes or high quality English teachers. 

Students come from homes where English is not spoken. Their parents cannot speak English or practice with them. Therefore the student app allows them to practice at home and give live feedback on their spoken English, even though no English is spoken. 

English teachers often struggle in the language themselves, and have received little or no training in how to teach English. 

Schools struggle to have good accountability systems for teachers or students, so they want regular data based updates on performance and implementation of the programme.

To date we have been funded by a mixture of revenue from customers, angel and venture capital investing, and grants. 

We plan to be profitable and self sustaining from customer revenue by June 2025. 

We are consistently increasing our revenue from customers, and decreasing our annual loss, and have a clear path to profitability by 2025. 

Our plan to be profitable by 2025 

In 2023, we made $226k and had costs of $647k, a loss of $421k,  making up the shortfall with angel investment, VC investment, and grant funding. 

In 2024 we forecast that we'll have $497k, costs of $799k , a loss of $302k, covered by our current cash balance of $350k. 

In 2025 we will make $1.345 million of revenue from customers, and incur $1.33 million in costs. 

Please see our financial forecast for 2024 & 2025 here 

https://docs.google.com/spread... 

How we are currently funded 

We are currently funded by revenue from our customers, venture capital investment, and grant funding. 

Revenue from our customers - we made $226k of revenue from sales to schools and other B2B organisations in 2023.  We forecast revenue of $900k for 2024. 

Venture capital - we have raised $1.6M in venture capital ($1M) and angel ($600k) funding. 

Grants - We received a grant of $100k from Emergent Ventures and $125k from ACT India.