Every few months, Yodle’s engineering department holds an internal hackathon that gives our developers a chance to take on an interesting project of their choice. This trimester there were four themes: UI Delight, Patient/Customer Experience, Artificial Intelligence, and Technology. Teams in New York and Austin formed to work on their ideas and then demoed their results after two days. When not busy working, engineers got to enjoy a retro Pac-man arcade tournament, paper airplane competition, episodes of Futurama, and dinner and a movie featuring the nostalgic 1995 movie “Hackers”. Congratulations to our winners!
UI Delight: Folding Card Animation
Lighthouse Field Service is a product made by Yodle for field service businesses to manage their operations and marketing. We wanted to reimagine the job list in our mobile app to highlight the most essential information for field technicians, including travel time. The start and end times are visualized in a vertical timeline to make them easier to read. Selecting a job expands into a detail view with a delightful folding card animation. This view was fully reimagined with a focus on a photo of the job location and quick actions for contacting the customer. We look forward to incorporating this design language in a future version of the app!
Customer Experience: Killing Me Softly
Users, whether they are our clients or internal employees, need to know the moment their experience may be impacted. With monitoring and alerting now engaging our engineering team within minutes – it seems silly to have to wait 20-30 minutes to get clear communication out to internal staff and customers. By connecting our Synthetic Transactions in New Relic, Zapier Cloud Automation, and StatusPage.IO – we can now notify users and customers instantly within the browser experience they already have. Instant notification means setting expectations and giving customers an acceptable experience while we address the underlying issue.
Artificial Intelligence: Lighthouse Predictor
A primary pain point for Lighthouse Dental clients is ensuring that patients will show up for their scheduled appointments. To assist offices with this, we’ve analyzed aggregate data for patients and appointments to predict whether patients are likely to complete future scheduled appointments. This information can be used to give offices context on which appointments they should take additional action on, and help ensure that their appointment book is as optimized as possible. Above is an image of the Spark cluster that was busy running a regression analysis to determine a predictive model. The model was able to predict whether a patient would miss an appointment with over 85% accuracy.
Technology: Kubernetes + IP Routing + Persistent Storage
A core group of Yodle’s Coolbernauts spent 2 days prototyping version 2.0 of Yodle’s Platform as a Service (while managing to also scour the internet for funny cat videos). The existing platform (v1) revolutionized how we deploy and orchestrate our application tier, enabling us to pursue a RESTful microservices architecture serving up our API and UI layers. However, the two big limitations that have come up numerous times involve: (1) deploying stateful applications (e.g. databases) and (2) non-HTTP service discovery and communication. Kubernetes (k8s) supports many of our use cases out of the box with minimal set up, and it makes it easy to implement the rest. The team spun up a new k8s compute cluster with DNS service discovery, and implemented the following on top of it:
- Canary Isolation (a cornerstone of our deployment pipeline)
- External Service Discovery and Routing of incoming HTTP Traffic
- Persistent volume allocations from NFS
- DNS Discovery via A and SRV records
We made giant strides towards revolutionizing our already state of the art platform as a service, and will continue to look into productionizing our work: supporting cross datacenter federation, integrating it with existing deployment tooling, and creating a migration plan from the existing cluster.
Crossing the Chasm: Auto Image Tagging
Image metadata can help people find images they are looking for and is a potential ranking factor for Google search. There are thousands of images across our clients’ websites without good metadata attached. We wanted to take on the challenge of creating meaningful descriptions of theses images. Obviously there are too many to do manually, so we used the new Google Vision API to tell us what objects are in each image. In our testing the Google Vision API had incredible accuracy, allowing us to automate a task that would have taken a significant number of man-hours not very long ago.
Highest Ratio of Votes to Team Members: Hey Lighthouse
Hey Lighthouse welcomes Lighthouse into the realm of voice recognition technology, recognizing user input and displaying information the user requests. Instead of clicking around searching for patients or appointments, Hey Lighthouse can do that for you! All you have to do is ask. Above is a screenshot of its responses to the questions “Show me my appointments for September 11th” and “Who am I seeing?”
Congratulations again to all of our winners! We had some really innovative projects this time around. We’re already looking forward to our next hackathon which we’ll be holding in just a few months!