HyperVerge - Powering Trust through AI

• When you ask any profitable company why they don’t reach out to the world’s poor to provide the same services they do to other income segments, they repeat an oft-used phrase- “the cost of operations is too high”.
• This is particularly true for affordable credit, and key parallel industries that enable income increment especially in poor households across the world like insurance, electricity, water or mobile phone networks and the internet.
• We provide a comprehensive suite of products that enable organisations to reduce their cost of delivery which includes operational costs and cost incurred on fraud applications.
• Economic growth is underpinned by the scale of transactions. Every second, dollar and ounce of effort we save, allows our client companies to transfer the gains of ease of access and affordable services to under-served households, eg: by lowering interest rates - and we can help save a lot of those.

• Globally, 1.7 billion adults don’t have a bank account, and over 190 million adults in India remain unbanked. The low-income segment is left out of the formal economy as they pose a higher risk of defaulting on debt. To combat this credit risk, banks employ tools like Identity Verification and Know Your Customer(KYC).
• For instance, to open a bank account, a low-income individual is required to submit a range of documents, and the lack of traditionally-accepted documentation makes the process harder. The complete KYC process can take up to 20-30 days if done offline.
• For a bank, the operational cost of serving low-income segments - setting up an account, getting to know their customer, and remittance costs - ends up being very high. This cost is shouldered by the user as ATM withdrawal fees, SMS alert fees, among others, and widens the access gap for low-income users.
• Apart from this, the cost incurred by a bank on fraud applications is also passed on to the user. For instance, 5-8% of all loan applications to digital lenders and 1-2% of applications to consumer lenders in Vietnam. are fraudulent HyperVerge’s Trust Solutions aim to reduce the costs of serving the low-income segment, and incentivize financial inclusion by paving the path for profitability.

• HyperVerge Trust Solutions include our customer onboarding stack and fraud check offerings.
• The customer onboarding solution is designed as a safe and private application process by simply requiring the customer to take a selfie and hold up their documents to a phone camera from their preferred location. Service providers with access to our technology provide their end users a completely digital KYC solution that reduces document collection time and onboarding costs.
• Our fraud checks flag suspicious cases during onboarding itself and additionally, we catch identity frauds by checking if a customer has provided the same identity details elsewhere in the past. This process is called deduplication.
• Our solution uses face recognition and liveness technologies to conduct face matches between selfie and ID cards. We also use our face search technology to efficiently search for a particular face through millions of faces in tens of milliseconds to perform deduplication checks. To decipher identity proofs and documents and verify their authenticity real time, we use our in-house Optical Character Recognition (OCR) technology clubbed with our quality checks & tampering detection technology.

The idea to build a solution that reduces the cost of delivery of financial and allied services has been slowly growing and developing in the company for over 6 years now. When the company was first formed, it was formed with the vision that a student-led Computer Vision Group in an Engineering College in India had the potential to build a better world through their competencies in AI. This group was a collection of individuals who had grown up in a low- to middle income developing country and were aware of the social and cultural challenges that continued to depress large sections of the society into intergenerational poverty.

At HyperVerge, we build B2B enterprise solutions in the Fintech and Geospatial industry. We work directly with a range of clients ranging from marque brands like SBI, Jio, Vodafone-Idea, Bajaj Finserv, FE Credit, KrazyBee to smaller fintech players and lenders. Together, this network provides financial solutions to millions of people across the world.

But the real possibilities of the impact of this work became clear to the company early last year. To take forward their larger goal of finding sustainable solutions for entrenched social problems, the company had set up a separate team called Contribution Initiatives in March 2020. The same month, India was hit by the Global Covid-19 Pandemic. The team organized volunteering groups to provide ration relief through local organizations like Reap Benefit. A program called "Adopt a Family" was also set up to give a zero-interest loan to some small businesspersons who needed instant capital to tide over the tough times without their livelihoods. The program enabled a lot of our employees to interact with families who had financial problems that went deeper than just the need for a small one-time loan. On the other hand, many of these families had outstanding informal loans where they were unaware of the interest rates, the tenure of the loan. They were simply stuck in debt traps they couldn't come out of.

To look for ways to address this problem systemically, the Contributions Team initiated two programs - The HyperVerge Academy (to upskill graduate students who find it difficult to find gainful employment and place them in well-paying jobs) and the Udhyami 250 Loan Pilot (which gave out zero interest working capital loans to revive small businesses after the First Covid Wave through a fully digital process). The goal was simply Income Increment for the family. The increment would give the household a chance to utilize their profits, to plan for better futures instead of simply barely breaking even.

The people, households and communities we work with typically earn less than $130/month. They either have adult members who have limited to no literacy or children who are first-generation high school graduates. They own small shops, rent carts to sell wares, are tailors and dry cleaners. Their livelihoods are informal, uncertain and precarious. They rarely have health insurance or life insurance. Many have only recently opened bank accounts and are unaware of the full scale of banking services. Proper documentation in the form of identity cards, or address proofs, is difficult to come across. Smartphones and feature phones are shared among multiple family members and digital literacy is even lower than formal literacy. Several social barriers disallow them from finding favour with formal service delivery systems offered by governments or private players. Their trust in institutions, as well as the trust institutions, place upon them, is low and stretched thin. Instead, most families find informal moneylenders they live in close proximity with more accessible than a bank branch.

With our partner organization, Udhyam Learning foundation, the Contribution Initiatives team has spent the last six months, researching existing models that enable giving easy credit to our target segment. With the Udhyami 250 loan pilot, our aim was to show formal credit lenders that such sections of society are trustworthy if the loan product is designed to suit their needs. As part of this project, we teach basic financial literacy and digital literacy skills to our target segment and prepare them to utilize their loans better. We call them regularly to check on their progress, understand problems and reflect on solutions we may offer them.

These have been some of our key learnings from the project:

1. Repayment Issues:

• Technological (Payment mismatch, Smartphone issues, Bank server issues),
• Personal (Health issues, death in the family, out of town due to emergencies),
• Business (low business, no income increment)

2. Financial Management:

Many entrepreneurs struggle with simple record keeping, calculating profit and loss and keeping track of expenses. Since there is little separation between business and home expenses, profits are low, vyapaaris are unable to invest back into their business. Most take repetitive loans to purchase inventory.

3. Operational Challenges:

Despite the project being completely online, there are challenges with onboarding, retention and engagement.

As part of these initiatives, we've also got a chance to interact with other organizations, companies, startups, community-based organizations, self-help group members who share their invaluable experiences of working with this target segment. These insights are regularly transferred and discussed with the larger HyperVerge team through weekly contribution Brainstorming Calls and Team Meetings. Many of our employees volunteer their time and language skills to translate our IEC materials to various Indian Languages. They volunteer to hold interactive sessions with our entrepreneurs.

These experiences have made us realize that we can create maximum impact by making a dent through our core competency as a company in lowering the cost of service delivery for the target segment. We are crucially placed at the crossroads with access to both our consumer target segment and the industry leaders in lending and service delivery.

These insights intrinsically guide the way we design our products. For example:

• Reduced turnaround time for customers: Marginalised sections of the society require funds fast for emergencies or other immediate requirements. By offering digitisation/identification services such as OCR, Face match and Liveness - with a click of couple of photos, we enable quick identification of the applicant for the clients and thereby reducing the amount of time it takes to process loan applications and thus disbursements to the end customer.
• Greater Supply: Technology is helping to onboard becoming easier - We are able to enable a lot more lenders to offer loans, which is better for the society as it gives them an option to choose among the many - this supply excess will in turn reduce the interest rates offered.  Moreover, this also means greater financial inclusion because of previously ignored or out of reach markets are being addressed remotely.
• Increased Access: Budget phones in tier-4 or 5 cities imply low-quality pictures. We develop top tech modules to work around this issue, accept, process and easily finish the application journey. This is critical for folks from rural areas to leverage the available technology. In Vietnam, our clients also offer local Kiosks where customers who don't have smartphones can access to upload their images and documents.
• Minimal Costs for the Customer: With the available digital-first companies offering loans, the amount of effort both in terms of costs associated with access, application have pretty much gone down to zero.
• Behavioural Impact: Digital lending offers a safe way for customers to seek finances. Customers who earlier considered it taboo to ask for money from the near and dear ones have a different outlet to get credit from now.  This is made possible by simply sharing a set of documents and getting them processed remotely using the tech tools that we offer.
• Literacy Requirements: Offering products online is becoming a smoother process than it is offline. Institutions use signs/cues to help even technologically challenged customers be able to apply easily. For example, we are able to help by drawing a box on the screen where the customer needs to place a document. The box turns red or green depending on whether the documentation is accepted. We also add culture-specific language cues.
• Small Businesses Adoption:  Plug and Play digital technology that we offer which can be integrated into systems within a limited time and easier integration efforts are leading to small lending businesses also adopting products that will benefit the business ecosystem and the people in the long term.
• Saving Time for our end user: Our decentralized fraud detection network proposed that the user only upload their identity once into the ecosystem. Whenever they need to register for any other service, the HV solution will pull the identity and provide it to the new service provider once the user gives consent. The best part is that the user is the owner of the data in this model.

Our target population and the challenges we face changes as the geography changes. For example, in Vietnam, we see the rates of fraud are much higher than India. ID tampering and identity theft is more common. We hope that as we scale this product we are able to engage directly with the target segment in each geography. But for now, we're dependent on our local clients to identify their demographics and its needs.

Finally, the AI solutions are agnostic to changes in facial hair, angles and lighting. It is compatible with older phones, tablets and web applications. It works even in Low bandwidth networks like 2G and 3G networks. It is optimized to handle racial diversity with minimal AI bias. These features affect our ability to build globally scalable products specially appreciated by clients who cater to millions like SBI and Jio.

We provide a scalable, safe, private and digital identity verification solution that can be used by financial service providers to extend their services to low-income communities and small businesses by reducing the cost of delivery.

• Safety & Privacy: Our solution is safe and private by design, as the ownership of the customer’s data remains with our clients and its end users. Having large financial organizations and telecom companies as clients, we ensure compliance with regulations of local government authorities.
• Scalable: Over the last 2 years, we have onboarded 450+ million customers across 5 countries and 100 clients.

Different parts of our solution are in different stages. With our identity verification stack, we have been able to onboard 450 million customers across 5 countries and 100 clients. This is in the Scale Stage.

However, the decentralized identity and data infrastructure are in the Growth Phase. We are still kickstarting our efforts in distributing the infrastructure we built for collaboration with 5-10 clients in India & Vietnam right now. We are now looking to scale that across customer segments & geographies.

Our solution rests on a four-legged approach to increase the overall impact of technology -

1. Improve accessibility
   • Our Liveness technology has helped clients onboard end-users by cutting down on the "English-tax" which penalizes users who do not speak English and would have to perform complicated gestures to prove they're real customers.
   • The UX design cuts down the requirement for digital literacy
   • We also use vernacular languages to ensure language is not a barrier.
2. Increase scale
   • Our technology can be deployed in low-bandwidth areas.
   • While our competition caters to end users with high-end devices (cost > $250), our tech supports up to 1 million device models (that cost less than $150 each)
3. Outsize impact
   • Our facial detection and recognition algorithms are non-intrusive and have an industry-leading accuracy of 99.5% to reduce fraud and consequently, reduce the transaction costs of doing business.
   • Our deduplication technology can parse millions of faces in milliseconds to check for fraud by verifying if the same identity details have already been used in the past - reducing fraud and the cognitive load on employees working on fraud checks.
4. A decentralized platform for collaboration
   • Government or other identity providers store all Personally Identifiable Information (PII) data centrally with them magnifying data security risks. By separating infrastructure, database & platform layers, we've made it friction-less for organisations to join the network while maintaining full control over their own database.

• On an average we serve 0.5 million end users on a daily basis
• So far, we have served a cumulative of 450 million users since 2018
• We are aiming to serve more than 200 million users in the next 1 year and 1 billion users in the next 5 years

Our direct impact metrics include:

• Ease of access, inclusivity
• Number of customers we serve: 450 million users in the last 2 years
• Money saved in detecting fraud: Eg- We estimate savings of upto Rs INR 1.7 Cr for a digital lender in India
• Money saved in onboarding cost: Rs 65/customer

But since we also check the impact our work has on our target segment, we will be measuring:

• Increase in total number of users (SDG 8.10, 9.3)
• Increase in new to credit customers (SDG 8.10, 9.3)
• Reduction in interest rates (SDG 8.10, 9.3)
• Demographics of users (SDG 10.2)
• Gender of users (SDG 5.a, 5.b, 10.2)

We are currently a 144-member team of which 2 of them are interning with us.

At HyperVerge we believe in "paying it forward"- to our families, communities and people. The company was formed in 2014 out of a student-led computer vision group from IIT Madras, India. As students, under the guidance of professors, the group executed projects for Indian Railways, ITC, MRF, etc. Since 2016, the company has pivoted to B2B enterprise solutions.

What brought the founding team together was a motivation to build solutions for entrenched social problems that the founders and early team members have seen all around them growing up in a low to middle-income country. With the goal to build an organization that contributes meaningfully to making people’s lives better, we started the Contribution Initiatives team in March 2020. This is a 5 member team with prior experience of working in the non-profit sector through education, social impact and public policy programmes.

Given the organizational proximity to financial services and technology*, the Contribution team set about to create impact through initiatives in financial inclusion and gainful employment. Through our zero-interest loan pilot and the HyperVerge Academy, we work with small businesses, young graduates and community-based organizations on a daily basis to alleviate their concerns. The insights we generate from these projects, help our technology and product teams make their products more inclusive.

Do watch this video to learn more about our story and about us as a team.

Our team is committed to creating a diverse ecosystem that discourages discrimination on the basis of race, gender, class, and caste. We are focused on developing long-term, sustainable solutions to counter the lack of diversity, particularly within tech.

One of our long-term initiatives is the HyperVerge Academy. The Academy tackles systemic inequity by training individuals from rural regions and underprivileged backgrounds in critical thinking, in-demand technical skills, language, and soft skills. These individuals also receive mentoring from the leadership within HyperVerge. While the Academy is less than ten months old, we have been successful in placing many first-generation degree-holders in leading organizations. One of the Academy's graduates found a position with a competitor.

However, the team at HyperVerge recognizes and acknowledges that we must do more to bridge the gaps. Although currently,

• three critical teams are led by women (our India Business Unit, the Marketing Team, and Contribution Team)
• other teams (Solution Architects, Talent Acquisition) are at least 50%+ female-identifying
• we have at least 50% local representation in key positions at international HyperVerge offices

We must do more to embrace inclusivity. As an organization focused on creating impact, we are continuously looking for more ways to build pathways for more diverse talent, including setting concrete goals to track our diversity initiatives.

• As a financially sustainable organisation largely in the scale-up stage, applying to Solve for funding is the last thing on our minds. We are very clear about our commitment to solving real-world problems that create a meaningful impact in society and we believe that being a part of the MIT network will enable us to forge deeper partnerships with the social impact community. Most importantly, we believe it will allow us to learn about the needs of various communities across geographies to make our solution even more inclusive in nature.
• While we have benchmarked our technology solutions against global standards and continue to do so, we are yet to benchmark ourselves from a social impact standpoint. We believe that MIT Solve sets high standards of excellence in this domain and it would help validate and recalibrate our understanding of impact.
• As an organization, we strongly believe in the principle of collaboration beyond competition. In that light, we are mainly looking forward to MIT Solve opening the doors for us to connect with a wide range of problem-solvers, social impact organisations, industry and domain experts we can partner with, to accelerate the impact we can create through our solution. We are already seeing the benefits of engaging with this community as we reached out to a few organizations (ION, YouVerify) who proposed a solution in the same category.

In order to ensure we take our solution to impact the most number of people in the shortest possible time, we would be grateful to seek help from the MIT ecosystem in the following facets based on our current strengths and requirements:

• Human Capital: To do justice to every major geography where this solution could create impact, we would need assistance from a dedicated local team and like-minded partner organizations to fast-track this journey. Apart from this, we require support from advisors who can serve as a sounding board and strategically steer the organization to avoid common pitfalls.
• Public Relations: In an endeavor to build the best solution that works across geographies, for varying ethnicities and bandwidth conditions, marketing was de-prioritized. However, to scale our reach and impact, we want to invest in active brand-building. Support from the MIT ecosystem in scaling up our marketing and public relations effort would have a disproportionate impact on us scaling our solution.
• Product/Service Distribution: While our primary work has been in the APAC market, there are numerous other regions in the world that are just getting started on their digital journey. Ensuring a robust digital identity framework across the globe entails building local partnerships who can further our mission of improving access.

We are looking to partner with organisations who can -

1. Help us tailor our solution for different geographies and domains. For instance, we would love to partner with:
   • Organizations like YouVerify, working on KYC and customer onboarding for people in Nigeria and other parts of Africa who are at a disadvantage due to lack of identity verification infrastructure that is keeping them from access to formal credit and other financial resources. We found them on the MIT Solver website.
   • Traditional banking systems providers like JackHenry who can help make our solution accessible to a cohort of banks at once and give us the ability to reach out to a large number of individuals, helping us create a bigger impact.
2. Help us evaluate the inclusivity of our AI engines
   • The Casual Conversations dataset from Facebook AI Research (FAIR) is the first public dataset for AI Inclusion. Working with similar research teams to help build unbiased AI solutions will allow us to rapidly expand across geographies, with a high degree of confidence in the fairness of our AI models.

HyperVerge is impacting financial inclusion by building the technology stack for financial service providers who cater to low-income communities. For instance, our work with a wallet partner in Philippines has helped to distribute COVID-related government subsidies to farmers and indigenous people. Specifically, our liveness and deduplication technology has sped up the disbursal and helped plug leakage by quickly identifying fraud. We believe our culturally sensitive technological solutions can rapidly accelerate the rate of access to critical financial services and deliver outsized impact.

The founding team bonded over a shared motivation to build solutions for entrenched social problems that the founders and early team members had seen all around them growing up in a low to middle-income country. HyperVerge's solutions, built over the last 4 years from the ground up, are enabling financial inclusion among populations with generations of inequity and helping them take their first steps in a world that is digitizing extremely quickly. Backed by powerful, intelligent AI engines that have won accolades on the global stage, we are truly delivering tomorrow's AI, today. We aim to serve 1 billion customers within the next 5 years and have continued to grow rapidly, even during the pandemic. With a stellar track record, solid industrial partnerships and a resilient, diverse team that infuses their humanity into every line of code, every interaction, we believe we are qualified to be recipients of the AI for Humanity prize.