Radics

Receiving consistent and personalised support from Parents, Teachers, and School Leaders is the cornerstone of student centered learning environments. Reliable diagnosis of student progress and communicating this diagnosis effectively is a key enabler for this support. But, widespread digital inequity  keeps low-resourced schools from practicing personalised student support. We want to change this with Radics.

Radics is a light-weight, AI-chat driven learning and communication management system that enables cost-effective collection of high-quality student data in online-offline settings via Chat, Mobile and OCR-enabled worksheets. Insights from raw data, enriched with evidence-based recommended next steps are shared with Teachers, Parents and School Leaders to personalise student support.

To assist behavioural shifts, Radics links usage and student success to a philanthropic support system called school credits - a blockchain-enabled reward system that converts positive shifts in student outcomes into  cash and in-kind support for essential school supplies and student scholarships.

Realising a student-centered support system rests on informed human interventions that are backed by evidence. Unfortunately, most schools in the developing world lack the necessary technology infrastructure and competence to engage in data-driven student support mechanisms. This keeps such schools from personalising instruction based on student strengths, emphasising social-emotional skill building and effectively involving parents in student support. The incumbent LMS and School ERP solutions that solve these issues fail to find widespread adoption due to unnecessary layers of complexity and high expenses. 

Experimental evidence from a variety of TaRL (teaching at the right level) studies suggest that the effect of devices and digital content is marginal as long as the student interventions can be personalised, esp. in K8 group. Because most schools in the developing countries can not afford to bring a device, internet and engaging digital content to each child, we need a cost-effective way to help these schools leapfrog from a conveyor-belt system to a student-centered system without having to invest in capital intensive blended learning infrastructure. 

How might we design a solution that can cost-effectively mobilise and direct the proximate human support systems for the child through lean yet powerful technologies?

Radics replaces traditional LMS and School ERP layer by a lean, AI-chat driven learning and communication management system that facilitates data-driven management of proximate child support by teachers and parents.

We radically improve the management of student support systems at three levels.

Sector - we systematically integrate a variety of views on student development and break the silos in scientific measurement of student development through innovative and open-sourced data instruments. These instruments codify both academic and non-academic outcomes, and can be implemented via a variety of mediums - chat, mobile app, worksheets.

School - we reduce the barriers to cost-effectively manage student data through a multilingual, AI-chat driven tool accessible for all school stakeholders. This tool distributes the measurement instruments, and automates data collection and stakeholder communication.

Stakeholder - we convert raw datasets into actionable insights for each key stakeholder i.e. teacher, parent and school leader. These actions are labelled and rated by a diverse set of practitioners, and are backed by strong student outcome evidence.

To drive behavioural shifts, we use a blockchain-enabled system to distribute philanthropic support for schools. Positive shifts in student outcomes are rewarded with donor-sponsored credits that can be converted to school supplies and scholarships.

Radics primarily serves Teachers, School Leaders and Parents - people who directly and most decisively affect the holistic development of a child. We design our solutions keeping in mind the digital divide faced by our target audience. 

We use rapid prototypes and A/B tests to identify the product fit and make continuous, weekly improvements based on telemetry data, and qualitative user feedback collected via surveys and workshops. 

An alpha version of our solution - Radics 1.0, built in collaboration with an education NGO, is in active use since Jan 2021 for 459 low-resourced schools across India, serving the learning management needs of 1910 teachers and school leaders, reaching 50K students and generating over 700K student test responses so far. Three key observations sum up our learnings from this pilot so far:-

1. Establishing the case for chat-driven learning data management

Chat presents a promising entry-point for establishing two-way communication with the most marginalised groups. Education too has seen chat being leveraged as a preferred medium to reach audiences that do not yet have the means to engage in rich digital learning activities such as HD Videos and Live tutoring - something our target beneficiaries confirm being devoid of. With Radics, we have early primary evidence in favour of chat as a viable solution to cost-effectively promote and practice hallmarks of 21st century learning i.e, personalised student support, socio-emotional skills and parental engagement. We have started our journey with a student data management tool (WhatsApp, 1900+ teachers, 50K+ students) and are continuously improvising chat as a viable medium to fill for the lack of popular digital learning solutions.

2. Proving that indirect ed-tech interventions can work and scale

Past decade has seen huge investments in edtech solutions that intend to replace teachers, and this trend is matched by teacher capacity and quality deterioration during the same period. However, as discussions around the relevance of social and emotional intelligence, along with emphasis on 5Cs gain momentum, the role of teacher is becoming central again. With Radics, we prove that keeping the needs of educators and parents at the heart of an ed-tech intervention can also gain momentum at-par if not better than solutions that bypass these stakeholders by design. In just over 6 months, we’ve observed more than 700K student data submissions on our platform where we have zero direct interface with these students, and depend entirely on teachers to disseminate content. This approach does slow our growth in the short run, but provides us room in the long-run to build technology that integrates well with the human support mechanisms for students. This approach is particularly reflective of how we address the digital inequity faced by our target student base and meet the need for proximate human support that is proven to drive excellence in student outcomes.

3. Finding a non-consumer base to disrupt the blended learning space

Strong non-consumption patterns are the bedrock of disruptive innovations i.e. patterns of early adoption by users for whom the alternative to an innovation is nothing. Blended learning is subject to similar disruption patterns in schools that haven’t seen much technology yet and lack even basic resources. However, because most schools are a closed innovation system with strong government oversight, it is difficult to find a reliable early adopter base from with-in the schools. This is particularly difficult for low-resourced schools. With Radics 1.0, we’ve found success in establishing a strong early adopter base in public and low-fee private schools, and aim to strengthen trust with our user base as we bring proposed improvements to our solution.

Further to our ongoing pilot, we are actively developing a beta version of Radics (Radics 2.0) with following new features:-

1. Statistically validated, open-sourced instruments to measure:-

   • foundational literacy and numeracy skills,

   • social and emotional well-being,

   • national curriculum based competencies (for India’s national curriculum framework),

   • teacher efficacy via classroom observations,

   • parent engagement and at-home support needs

2. Evidence-based repository of micro-actions that have strong correlation to student academic and non-academic development. These micro-actions will be triggered by availability of new data insights, and be served as nudge chat messages to teachers, parents and school leaders

3. Donor-backed credit system to help schools make improvements to their teaching-learning capacity, including but not limited to bridging the digital divide. Additionally, schools could use these credits to award student scholarships

Radics solves for pain-points faced by teachers in figuring out where each student is in terms of academic achievement and non-academic support needs. We simplify the entire process of collecting regular and reliable student datasets, and then leverage machine learning to break these datasets into actionable insights. Teachers use these insights to:-

1. Provide personalised support to each student or a group of students,

2. Nudge parents to take specific steps for better at-home learning environment, and 

3. Collaborate with school leaders to design school-level interventions such as remedial classes

Our aim is to become a viable and more scalable alternative to incumbent LMS and School ERP tools. This is a complex task because improving formal learning structures demand close coordination with the human layer of student support esp. teachers, parents and school leaders. As we’ve put our bets on AI-chat as an entry point, we’ll need a certain degree of interaction with these human layers to deliver a reliable solution - something we are actively doing with our prototype. 

Further, as we've generated significant datasets over the last 6 months, we are now in a position to design the proposed ML-enabled data analysis techniques for making the data practically useful. As we improve our analysis capabilities, we'll also implement the proposed incentive mechanism ie school credits to effectively move to a pilot phase.

Radics flips the learning data management problem from being an organisational concern to a consumer issue. Most learning data management solutions start by defining what school administration needs as an organisation as opposed to what the individual stakeholders’ pain points are. Meeting students, teachers, parents and school leaders where they are can have a significant impact on how they interact with technology and use data. Adhering to this principle becomes further challenging for low-resource settings that introduce additional constraints related to language, tech literacy and online-offline data synchronisation. We innovate on four fronts to address our specific challenges and practice a consumer mindset.

One, users get to interact with our system through a conversational interface that’s personalised for their specific needs. 

Two, we complement online-chat with low-inference, tamper-proof data collection strategies such as OCR-enabled worksheets, digitised using AI image processing in order to include the most marginalised students. 

Third, we leverage machine learning to break raw datasets into actionable insights that have a known and proven impact on student outcomes. 

Fourth, we leverage blockchain to put student data to further use and build donor trust to mobilise tangible resources for school improvement and student support.

We believe that our approach could potentially pave the way for incumbents and newcomers alike to focus on building contextual technology solutions, leverage AI/ML to tackle unresolved barriers to imparting quality education in low-resource settings, and further the use of blockchain technology in education.

We currently serve 1910 teachers and school leaders from 459 schools. These teachers and school leaders collectively meet the learning needs of 49K+ students. 

In one year - we'll on-board 5000 under-resourced primary and upper-primary schools, reaching 500K students

In five years - we'll on-board 100K under-resourced primary and upper-primary schools, reaching 10Mn+ students

Impact Metrics

Impact Objective: Gains in academic and non-academic student outcomes at par with TaRL (teaching at the right level) and related experimental studies. Focus on:-

• Foundational literacy and numeracy
• National curriculum prescribed competencies in Maths, Language and Sciences
• Social and emotional skills

Our MnE strategy to achieve the intended impact can be understood by reviewing our Mar 2022 Objectives and Key Results shared below.

Outcome Metrics

Outcome Objective: Under-resourced schools practicing student centred learning

Key result: 5K Schools, 500K Students by Mar 2022 actively using Radics platform

Output Metrics

Output Objective #1: Reliable, easy-to-use offline-online data collection tools

Key results: 

• At-least 50% of the entire student-base submits weekly/bi-weekly data irrespective of digital divide
• At-least 25% of the entire student-base submits monthly data irrespective of digital divide
• Maintain healthy NPS (9) and Avg App/Chatbot ratings (4+)

Output Objective #2: Holistic and statistically validated open-sourced measurement instruments

Key results:

• Tests for foundational literacy and numeracy in at-least 3 languages
• Tests for syllabus prescribed competencies (G1-G8) for at-least 3 subjects and 2 school boards
• One reliable instrument to track SEL and student well-being
• At-least 3 strategic partnerships to build and promote the open source community

Output Objective #3: Practical, actionable insights with strong correlation to student outcomes

Key results:

• At-least 100 evidence-based micro-actions linked to academic and non-academic student outcomes

Output Objective #4: Outcome-linked financial incentives for schools

Key results: 

• Mobilise philanthropic resources (cash + in-kind) of at-least $500K to be awarded as school credits

3 in core team (2 full-time, 1 part-time)

2 contractual engagements

Our core team comprises of three passionate Teach For India alumni who also have deep technical and design expertise to execute the proposed product vision for Radics. With a successful prototype we've demonstrated our ability to effectively meet our target users where they are and implement a software system that leverages a microservices-based architecture and basic AI/ML techniques, with minimal capital requirements.

Our team lead, Anand, brings over 14 years of education and technology experience with him. He has first-hand experience of the issues faced by teachers, school leaders and parents from under-resourced schools through a two year, full-time teaching job he took up in a Delhi slum area as a Teach For India fellow. He later joined the leadership team at Teach For India, heading the technology department and helping the organisation build student assessment and teacher development systems. Apart from Radics, he actively consults with World Bank's education group for large scale, technology-enabled ed reforms in India. 

Amit, who leads our AI/ML and content efforts, is an engineering graduate and a trained data scientist. Amit also is an alumni of Teach For India fellowship program where he taught secondary grades in a low-resourced school in Mumbai for two years.

Anshul leads our UX/UI and blockchain efforts. He brings with him over 5 years of applied UX/UI where he designed novel ed-tech solutions for a set of NGOs in India, including Teach For India. He is also a blockchain enthusiast and actively explores practical implementations of HyperLedger and Etherium.

Our current core team represents diversity and equality in light of varying degree of experiences. It is this right mix of experience and youth that helps us stay on course and implement complex product visions. However, our current core team is a product of years of shared experiences and  therefore may not be a perfect example of a team as imagined in MIT Solve's DEI statement. 

As we gain momentum and go beyond a fledgeling startup to become a serious edtech company, we intend to expand our core team to have representation from women in tech and education, and student leaders. We are continuously learning and growing, and building a diverse, inclusive and equitable team will be an essential part of that.

Problem-Solution Fit: we strongly resonated with the challenge description, especially the needs articulated for teachers. As we submit our formal MIT solve application, we want understand whether and how are we aligned to this challenge from an evaluator's perspective  

Visibility and MIT Solve network: as winners of a partner challenge we have had some exposure to how MIT Solve platform can increase one's visibility, and lead to support and connections crucial for the success of a fledgling start-up. By solving for a global challenge we intend to further our support ecosystem

Funding: grant and equity form a critical component of our sustainability strategy, esp as we aim to make heavy upfront investments in new product features and customer acquisitions. We are participating to win this challenge so that we can fund our growth over the next 2 years.

Human capital: we are looking for onboarding new talent in learning sciences, marketing and blockchain domains

Business model: our experiment with a direct to schools approach begins ion Sep 2021. Constructive feedback to devise a better gtm strategy will be really helpful at this stage

Public relations / marketing: we have operated in stealth mode, working as a backbone provider to other organisations. As we redefine ourselves as a B2C/B2B2C edtech startup, we'll need expert advise on marketing and branding matters

Product distribution: leads for new client and/or collaboration engagements to expand our client base will be really helpful too

Product Design and Development

We'd like to partner with education researchers, and non-profits that are looking to develop innovative tools for measuring student development. We bring on-board our own experience in this field, and a platform to statistically validate these instruments before scale-up.

Additionally, we'd like to collaborate with MIT Media Labs (esp. the education division) to conduct a critical review of our solutions architecture, and recommend ways to improve the same

Business Model and Funder Relations

We look forward to work with the MIT Solve team to hone our business model, and convert potential funder synergies into long-term, strategic relationships

With Radics 1.0 we brought Science and Maths digital learning material and practice tests to 450+ public and low-fee private schools in India. As we develop the proposed improvements in our platform, we'll maintain our strong focus on STEM education. Our primary value add is to help teachers and school leaders with the necessary intelligence and resources to improve the quality of instruction through personalisation. Our resolve to develop open-sourced and reliable measures of student success in STEM will help us deliver this value add to not just our client-base, but any other organisation, government or researcher who intends to improve the delivery of their own STEM programs.

Natural language processing

A large majority of our current user base are first time adopters of learning data management systems, and primarily interact with our solution though a  vernacular medium conversational interface. To that end, we've proven the efficacy of our natural language processing layer for basic interactions involving educational content access and student data analysis. As we progress towards building a reliable niche AI that understands the nuances of teaching-learning sciences, we aim to make significant investments in improving our NLP capabilities.

Image processing

Our pilot revealed a significant digital divide between rural and urban schools (student responses from urban settings outweighed those from rural settings by a factor of 6:1). To address this gap, we are actively working to leverage image processing technology that can simplify data collection from no-internet settings. For this, we plan to deliver student learning and assessment instruments as ocr-enabled worksheets that can be digitised with minimal manual intervention.

Machine learning

Our analytics layer will leverage machine-learning to convert raw student datasets into specific teacher, parent and school leader actions that are evidence based and have a proven impact on student outcomes. We will be training this action recommender with the help of practitioners, and will leverage ML algorithms to mimic the expert behaviour.

Open IRT and Data Science

Our open sourced measurement instrument repository will be based on a combination of item response theory, and deep learning techniques for statistical verification of each question item, and question sets.

Natural language processing

A large majority of our current user base are first time adopters of learning data management systems, and primarily interact with our solution though a  vernacular medium conversational interface. To that end, we've proven the efficacy of our natural language processing layer for basic interactions involving educational content access and student data analysis. As we progress towards building a reliable niche AI that understands the nuances of teaching-learning sciences, we aim to make significant investments in improving our NLP capabilities.

Image processing

Our pilot revealed a significant digital divide between rural and urban schools (student responses from urban settings outweighed those from rural settings by a factor of 6:1). To address this gap, we are actively working to leverage image processing technology that can simplify data collection from no-internet settings. For this, we plan to deliver student learning and assessment instruments as ocr-enabled worksheets that can be digitised with minimal manual intervention.

Machine learning

Our analytics layer will leverage machine-learning to convert raw student datasets into specific teacher, parent and school leader actions that are evidence based and have a proven impact on student outcomes. We will be training this action recommender with the help of practitioners, and will leverage ML algorithms to mimic the expert behaviour.

Open IRT and Data Science

Our open sourced measurement instrument repository will be based on a combination of item response theory, and deep learning techniques for statistical verification of each question item, and question sets.

Student portfolios as a viable blockchain use case has been around for a while. However, the same has been limited to facilitating school to college or college to work transitions. What if we could map student outcomes as a blockchain asset, and use the same as an immutable and reliable proxy for mobilising and distributing philanthropic support? With the idea of school credits we seek to answer this question.

We have some experience running a non-blockchain based resource exchange for under-resourced schools through a district-wide collective. However, trust was the central issue as we struggled to convince donors and meet our fund-raising targets. By introducing a blockchain to guide the distribution of philanthropic capital and in-kind donations we plan to address the trust issue among donors and schools. Radics is strategically placed to test this hypothesis as generation of high quality student outcome data is integral to its design.