Artificial Intelligence For Property Insurance
Editor’s note: Fresh off the heels of a successful Seed round, the founders of this thriving startup take a Medium pit-stop to share their vision
Tensorflight uses these principles — and beyond — to analyze satellite, aerial, drone and Street View imagery to make the process of insurance inspections less time and labor-intensive and thereby cheaper.
Our business is helping insurers and insurance customers with faster, cheaper inspections.
You recently closed your Seed round — congrats! How will the investment help grow your business? The new funds will help us scale sales and our computer vision algorithms to work with a global data set.
That work enabled me (Zbigniew) to co-author Inception, the most famous and influential deep neural network structure, which got 300 citations in just 1 year. It was the first computer model to beat human accuracy on ImageNet.
Now my vision algorithms also recently beat the world records in the most important computer image recognition benchmarks: ImageNet classification and MS COCO.
Each role I had at Google (Robert) helped me hone my skills.
From ranking financial products for Google Compare using machine learning to processing terabytes of data for Google Search, working with quantitative hedge funds or working with Artificial Intelligence and machine learning — every project helped contribute to what we’re doing now.
What’s been your best and worst startup experience to date? The best experience has been having the freedom to code! We’re no longer held to the constraints of following a set process.
The downside is fundraising. It basically amounted to two months of begging!
Are there any lessons from your time at Google that have helped your business? Our combined experience with deep learning, scalability, machine learning, the insurance industry and big data processing have all helped.
Each project at Google helped shape opinions about our new industry and guide the business decisions we’ve taken with our startup.
A few of our friends at Google Cloud also helped us with Kubernetes, an open-source system for automating the deployment, scaling, and management of containerized applications. They also helped us with Google Cloud Credits and beyond.
Two friends at Google Maps helped us understand the API and imagery partnerships. Thanks to Zbigniew’s experience with Street View we were also able to conceptualize how to integrate this kind of data into our business model.
What does it mean to you to be a “Xoogler? To continue to represent Google even as an ex-employee, keep strong connection with friends there and show everyone the best startups are founded by Xooglers(!).
How can the Xoogler community support your startup? Reach out if you can help us with credits, Kubernetes or other useful features that support deep learning startups.
We’re also looking for help with the Google Maps API, particularly accessing it and keeping up with new features. Similarly help with Street View, Google Places and the geocoding API are welcome. We’re always looking to add new features from Tensorflow e.g. how to build a good instance segmentation model.
Ultimately we’d like to build partnerships with Google to integrate its vast repository of data, especially maps and street view, into our product to help analyze it and provide value to the insurance industry.
Thanks both for your time!