HyperVerge- The game-changer in deploying a fraud detection solution at scale using cutting edge AI
In this podcast, we spoke with Kedar Kulkarni, CEO of HyperVerge a company that began in 2014 as a Consumer AI product company that went for a crowdfunding campaign but managed to raise a few million dollars in funding. When they were two months away from the product launch, Google Photos came in and essentially made the product being developed by them useless. There was a chance to basically sell the company at a high valuation and go for an awesome exit but they decided to persist as the idea behind starting the company was to solve a problem.
When there was about $500K left in the account, on their soul-searching journey to transform the company they meet Sridhar Vembu, Founder of Zoho who gave them the direction and deep-diving into the pieces of technology which was built. With the help of mentors, the company pivoted to a leader in Digital KYC and is now working on Video KYC.
In a place where the culture is built to treat Humans as capital and not a resource, the onboarding is a long process as the person joins the team to be part of the long-term journey.
It takes a village to raise a startup and its initiatives to upskill the people so that they can raise the economic standard of their families, and during the Covid pandemic, the financial and mentoring support provided to nano-entrepreneurs touches the heart.
Inspired by Batman and the learnings from Isha Leadership Academy gives a glimpse and insight into how passionate people with a vision to create an impact come together to build a thriving organization that solves a genuine challenge and, in the process, impacts the lives of millions of people such that during the pandemic, a client which had an in-house team to take over the work being done by HyperVerge as a vendor chooses to work with HyperVerge.
Theme Introduction: In this podcast series, we will have a light and casual chat with Founders, Tech Leads, Hiring Managers/HRs, Data Scientist, Researchers, Community Builders about how they use AI to solve real business problems. This podcast will bring you closer to all these amazing companies and get you excited to work with them. So don’t forget to stay tuned to all the upcoming episodes of the series and make sure to subscribe and follow us.
Ideation Days
IIT Madras used to be represented as Institute in India while representing its Robotics competition.
A bunch of students was participating in a Robotics competition each year that was transformed into a Research Group which worked on challenging problems from AI/ML, and there is no certificate, no recognition for this activity happening at a lab called Centre for Innovation.
Each year hundreds of students would join in, and at the end of it, there will be only a bunch of 4–5 passionate people who absolutely enjoyed working with each other on technology.
It was like a family, and after four years, the resurgent group began and the respect for each other.
In 2014, the Hyperverge team from college went to the US to demonstrate the technology as a Research Group with an aim to raise $50,000/- from a Kickstarter campaign, but people loved the tech in consumer AI so much that they ended up raising millions of dollars in funding and for a couple of years building out Consumer Apps.
Development Days
All of us take photos and we want to have a way to search our old photos, say photos with a friend or on the beach. That’s where the Picsaur (Dinosaur of all Pictures) app was born.
AI using Face Recognition, Scene Recognition could be used to sort it out on a platform where people will upload their pics and would be able to Search their photos. Like Search Engine for photos and nine months into building the product after funding.
In 2015, Google photos were launched, and Hyperverge’s product “Silver” was two months away from launch. Google Photos had entire functions available for free as the default app on millions of Android phones.
There was a pivot in doing another thing that could help users, but it has hit the wall.
Now a solid technology was built for Face Recognition with high accuracy engines that can work on very low bandwidth with very high scene recognition.
Most of the apps built were not being adopted as there were no 500,000 downloads or sort of million downloads and there were not many ideas on how to scale it beyond.
Key Learnings
Whatever you are working on, ensure that your team is always working with the highest possible intensity. Many times you will hit a brick wall. If you do so with a high intensity you will hurt yourself but learn everything you needed to learn from that experience.
If you do not go with the highest possible intensity, you will never know if your failure was a result of insufficient effort or a genuine reason that you had not known before.
Acquisition Days
In May 2016, there was some inbound interest for acquisition from top global tech companies of the world for the high-quality technology which was built.
Going down that path would have been like a double-digit million-dollar acquisition. It could have been a good outcome.
Starting a company that gets acquired for a budget is not a battle; the journey ending this way was not what had been set.
I took a few days off to understand what kind of work we may do even if we are not getting paid. Then came a Realization that we want to work to contribute in a significant way to society.
How do we go about doing something right with 6–7 months of money left in the bank account?
Defending Days
With 7 months on the runway, you could not significantly contribute to people’s lives, and before we know it, the money is withdrawn.
And during the journey very profound friends were made and when you are selecting your team you are creating a foundation and while reinventing the shift is towards creating products that would impact a lot of people.
Immediately, the team came together, identified the idea, and huddled together to develop the product called Digital KYC.
After meeting Sridhar Vembu, CEO of Zoho it was bought out that if you really want to contribute to people, you need to first learn how to make your own money, a way to survive and if you can survive you can also spend time contributing to life or not.
500k is the money needed in the first year to survive.
Where will that money come from?
Sridhar said some parts could come from a solution.
And one thing was clear Zoho can be the first client and start making an effort in that direction.
Now some serious questions to be very clear.
What will make us money?
What will you get as a Product Market Fit?
The team had no clue.
Prosperous Days
But there were pieces:
Wherever people are processing 100 million-plus images, the skills are valuable because there is a huge difference in something being 90% accurate, 95% accurate, or 99.5% accurate.
Imagine 10% better in 100 million is a huge number of errors happening.
I went after 8 different problems,
Looked at different sectors.
Looked at manufacturing, ATM, and Security kind of solutions.
Automation of Traffic Management, fines, and so on.
They got connected to Kalidas Ghosh, CEO of FE CREDIT, a FinTech company, and explained the capabilities and technology which has been built, and he could see its application in his business in the KYC process, which was Customer Verification.
Building the product, 60–70% of the learning came from the client.
What is the thought process in KYC?
How does the KYC process happen?
Finding each fraudulent case could save $1,000 or $2,000 for them.
A huge part of learning and product-market fit came from the client.
Every time you open a Bank Account, apply for Loan, Credit Card, and many other functions.
These functions are prone to fraud data like a person getting a loan with fraudulent documents, and they can run away with the money.
Get a real Sim on a fake account, and you can do many things with it.
These places are high risk and high motivation for fraudsters to break in as they can make money if they break through.
And while making this product, the learning was on the job.
While working with clients or their partners they looked at the kinds of frauds which was faced.
What are the kind of challenges while facing fraud?
Whenever there was a manual process, a limited percentage of fraud was caught and this learning leads to building AI modules that started detecting 90% of the fraud and eventually 95% of fraud, and now the number is much higher.
Similar learning comes from Claims, what kind of frauds are received over time, in a specific geographical area, and over time you have a foolproof solution as your defenses become stronger.
There are internal teams trying out many experiments, some parts are anticipated successfully, some are predicted and in a matter of a few days, the gap is plugged.
Like creating a new piece of technology which that stops them from getting through, like a Cat & Mouse Game.
Last year, 280 million cases were processed across multiple clients in India.
For a scale like this, the technology hardware that goes behind this plays a big role, and with the kind of infrastructure needed, the “On-Cloud” solution would provide scale and flexibility for working on 280 million images.
The AI engines have to work on images captured by a basic phone on a 2G connection to the latest smartphone across the country from rural, semi-urban & urban populations, ensuring no biases are crept in across approximately 100,000 unique devices.
Let’s have some fun moments with Kedar,
Social Days
In solving the real-world problem, Education is close to the heart which started in 2010 while at IIT working in Lema Labs teaching a lot of students and that’s where they meet with the Co-Founder. At Lema Labs, the education related to Machine Learning, Robotics, and Emerging Technologies was available to every middle-class student for a month’s salary earned by their families, and that’s the way it was learned.
Now that HyperVerge’s survival was figured out simultaneously, the plan was to help some families escape the cycle of poverty through education.
The families have kids aged 20–25 years, and they currently have BPO jobs or some other employment source where they earn about Rs. 15,000/- and while working if the person is able to learn For Eg. Front End Development. The same family income would rise to Rs. 30,000.
Now working with the first batch of students to help them come out of the cycle of disadvantage very soon.
It takes a village to raise a startup or a child, and we come from a middle-class family where we have learned a lot from society. As the Covid-19 pandemic hit, many people were without a job, and thankfully, the impact of the industry in which we are working was minimal.
Many people were rendered jobless, even with the government supporting them with limited funds and insufficient rations, so they decided to adopt 10–20 families by giving them Rs. 1,000/- every 15 days so they can get through.
While working with them we realized that some of them are very entrepreneurial and by giving them the basic funds they will be able to set up their 1–2 person business and sustain which led to our nano-entrepreneur initiative.
As the first lockdown ended, a lady needed Rs. 7,000/- to start her business, and with that amount, she went and bought a few ladies’ clothes and was able to sell them in her locality and able to sustain and grow her business.
An autorickshaw driver used to sell vegetables before and with Rs. 10,000/- as an investment, he could source vegetables from the market and sell them as he knew what had to be done to be successful, and with this boost, they were thriving.
With this initiative being successful, where we are supporting 20–30 families, we have partnered with an organization to figure out how we can support 2,000 families. It’s simply Pay-It-Forward as we have thrived on having great mentors in our life to guide us selflessly, be it IIT professors, Alumni, Clients, or society.
Hiring @ HyperVerge
Listen to the full episode here 🎧
If you liked what we did and want to know more about our community 👥 then please consider subscribing, following, and joining it. It is completely FREE.
Also, don’t forget to show your love ❤️ by clapping 👏 for this article and let us know your views 💬 in the comment.
Join here: https://linktr.ee/colearninglounge