bolttech: Seamless user experience thanks to next-gen machine learning
Driving digital innovation with AI and computer vision to deliver a pioneering new business model.
At a glance
DiUS helped bolttech enable a new kind of customer experience. Using pioneering remote diagnostics technology, it quickly and easily onboards customers onto device protection plans. Customers simply hold their smartphone in front of a mirror and move through a sequence of tests, powered by next-gen machine learning and computer vision technology. The result was a zero-touch risk mitigation tool for bolttech and a best-in-class experience for customers.
Based in Singapore, bolttech is an international insurtech with a mission to build the world’s leading technology-enabled ecosystem for protection and insurance. Insurtech, short for insurance technology, is a term used to refer to technology designed to enhance the operations of insurance firms and the industry as a whole.
As a digital native with a global reach, bolttech operates in 30 markets across three continents, North America, Asia and Europe, serving more than 8.3 million customers. With a full suite of digital and data-driven capabilities, bolttech powers connections between insurers, distributors, and customers to make it easier and more efficient to buy and sell insurance and protection products.
The team at bolttech wanted to create a new, AI-driven experience for protection and insurance for used devices — replacing a process that typically required a physical inspection with a fully digital solution. With remote servicing options for customers accelerated by COVID-19, bolttech wanted to continue supporting customers through digital experiences that are just as valuable as face-to-face interactions.
Looking for a technology partner with world-class expertise in machine learning to help bring the bolttech’s vision to life, the team enlisted the help of DiUS. The brief was to develop a computer vision model that would deliver a seamless customer experience, and could assess and classify the condition of the device (e.g. a smartphone screen—and whether or not it’s cracked or damaged) — on the fly.
What we did
Harnessing the power of machine learning
Working closely with bolttech’s innovation lab, DiUS helped develop a solution that leveraged the state-of-the-art, open source image classification model, FixResNet—released by Facebook just a couple of weeks before the project kicked off—run in PyTorch and hosted on Amazon SageMaker, in just six weeks.
DiUS was able to leverage our prior expertise in computer vision to help accelerate the machine learning model development. For example, one of our startups, Datarock, is focused on delivering important geological information about a mining site through machine learning algorithms that process digital photos of geophysical measurements from drill core trays.
In parallel, the DiUS team helped improve the data collection process to support model training. Then, to get the performance needed, the team trained the model by applying data-augmentation and multi-GPU model parallelism. Overall, we achieved a massive improvement in model accuracy versus initial prototyping, with zero-downtime in deployment of model updates.
To obtain an image of the smartphone screen for the purchase journey, DiUS built a model using Google’s MobileNetV3 with quantisation-aware training to guide the customer through the experience.
Because the model needed to be easily downloaded and run in real-time on a customer’s smartphone, a light-weight model was developed, less than 1MB in size. This was achieved by utilising quantisation-aware training which leads to models much smaller in size, with minimal impact on the performance.
Results for bolttech
bolttech’s easy-to-use diagnostic tool for device insurance and protection is called Click-to-Protect—a fitting name for a product that takes minimal customer to purchase effort, digitally and seamlessly. It’s a computer vision-based screen diagnostic that helps bolttech to understand device health when offering device protection for a wide range of devices.
The final experience directs the customer to hold their smartphone in front of a mirror. It then guides them through feedback coming from the light-weight machine learning model, hosted on the edge device, in order to take photos from the front camera of the smartphone. These images are then analysed by the backend model to ensure the smartphone screen is uncracked.
What would have traditionally taken a number of forms and a physical inspection, now only takes seconds to complete. The new approach not only reduces the cost and effort of setting up new coverage plans, but it creates a vastly superior digital experience for bolttech’s customers.
Click-to-Protect is currently performing at accuracy levels well in excess of targeted benchmarks. The model is continuously re-trained with libraries of images to improve the performance.
Click-to-Protect has gone live in seven markets including France, Hong Kong, Italy, Malaysia, Philippines, Thailand, and South Korea, with eight partners in total.
bolttech’s footprint is rapidly growing, and so are its capabilities because Click-to-Protect is expected to launch soon for smart watches. Further development is focusing on other enhanced capabilities such as fraud detection via OCR, QR, device fingerprinting, and further improvements to the automatic object detection model.
“Delivering a production-grade, computer-vision driven customer experience needs more than technical expertise. It requires a deep understanding of the way humans interact with machines and in how to train machines to understand and process real-world data and images.
DiUS’ proven expertise in AI and learning-based image segmentation analysis helped us rapidly progress from proof of concept to production, to continuous delivery. This has been executed with improving model accuracy, accelerated training time and zero-downtime in deployment of model updates.”
– David Lynch, Group Chief Technology Officer, bolttech.