Data store migrations with no downtime

Data store migrations can be complex, especially if the application cannot afford to have maintenance downtime. We will explore some strategies that can help you navigate the complexity. In our case, it was a migration from a NoSQL data store, DynamoDB into a relational data store, MySQL. However, this post…

Rotational Security Interns

The internship will…

How we improved the observability of a Go project

Background About a year ago, my team took over responsibility for an event driven system written in Go. One of our first steps was to understand what the application does, and how it does it. The app already had stdout logging, rollbar notifications, datadog statsd metrics, and pager alerts. The previous…

Machine learning is blossoming at Zendesk

Bill and Ted’s Excellent Adventure! A Short History of Machine Learning at Zendesk We have invested heavily in ML over the past four years and we…

Code Coverage and Xcode

Key to unit testing is code coverage…

Building user confidence in machine learning products

I’m a Software Engineer on the Data Products team at Zendesk. Recently I had one of the highlights of my career — working in the team that built Answer Bot. As a front-end specialist, my focus was on crafting the best possible user experience for Zendesk customers…

The science behind consolidating Answer Bot production Models: Part 1

At Zendesk, we believe that great customer service builds strong customer relationships and that Machine Learning ( ML ) can play a powerful role in creating good customer experiences. Whether it is enhancing search tools or powering entire self-service systems, ML has the…

Datadog APM Trace Search from Zero to One

Trace Search and Analytics by Datadog HQ is a fantastic tool for performance analysis and capacity planning. Trace Search has been in Beta over the past year and was announced for General Availability at Dashcon.io. Please see the keynote, related workshop and also its Github repo. In this blogpost we…

The Joys of Circuit Breaking

Zendesk’s product suite is built with an evolving service oriented architecture, comprising hundreds of micro services and a growing web of interdependencies. Operating and orchestrating such a large scale distributed system can be challenging. This post will explore the importance of circuit breaking and related concepts. A few weeks ago…

A Beginners Guide to Deploys and Releases

The process of releasing software to users at scale is one that can be particularly confusing for new software engineers, as most don’t have exposure to this part of the software development process until their first internship or job. I found this to be the case during my time at…

Part 2 — Zendesk Spotlight: Our Career-Changer Engineers.

Following on from Zendesk Spotlight: Our Career-Changer Engineers I received so many words of wisdom from our career-changer engineers that it was clearly worth sharing it with the world. And judging by your response to Part 1 you think so too! Here is what they had to say. Catherine Software Engineer…

Performing Kubernetes Service CIDR Block Migration

History At Zendesk, we started our Kubernetes journey in September 2015, shortly after version 1.0 was publicly available. At the time of v1.0 not much information was available and we didn’t have a lot of experience, so some configuration choices that made sense four years ago have had to be reconsidered…

Leading Machine Learning at Zendesk

I recently participated in an interview series covering careers in emerging technology fields on the French website TOM.travel. I’ve been leading machine learning at Zendesk for the last 6 years and over that time we’ve grown from two people in Melbourne to tens of people on three continents. I discussed…

Zendesk’s Global Mesh Network- Part 1

Running a team colocation — The TL;DR

NO to over-planning, YES to #COLOYOLO Don’t over-plan and take it easy! Make it a working week, don’t try to overload it with meetings. Just remember…

The science behind consolidating Answer Bot production Models: Part 2

In the previous post (Part 1), we introduced the issue of consolidating the six production models that are used to provide multilingual support for Answer Bot at Zendesk into a single model. Consolidating these models both simplifies our deployment process for new models, while providing a potentially more scalable solution…

Going Where our Team has not Gone Before

How to plan out where you’re going when you don’t know how you’re going to get there. Organising a data science team’s research direction is a challenge in itself and there’s no algorithm with which to optimise it. This quarter our team is trying something new. Generally when our data…

Chaos Engineering comes to Ruby

In a micro-service environment, an unexpected service outage often comes as a surprise and when it happens it’s stressful for engineers, managers, and the clients. If you want to prepare your Ruby application for the next surprise and prevent it early then this post is for you. I will show…

A minimal Python client for talking to TensorFlow Serving

min-tfs-client is a minimal gRPC client for TensorFlow Serving that doesn’t depend on TensorFlow. TL;DR: min-tfs-client is a minimal python client for TensorFlow Serving that allows you to use serverless services (e.g. AWS Lambda) that have deployment size limits (250 Mb uncompressed, at time of writing). It works by removing TensorFlow as a dependency to create tensor protobufs for the prediction request. Feel free…

The history of infrastructure at Zendesk (Part 1) — constant tradeoffs

Maintaining Binary Compatibility in Scala

Zendesk Spotlight: Our Career-Changer Engineers.

(Part One) Zendesk Engineering in Melbourne is growing, and growing fast. Four years ago we had five engineering teams and now there is somewhere in the vicinity of thirty. One of the pillars of this rapid growth has been the hiring of career changers — folks who have decided to shift their…

Zendesk ML Model Building Pipeline on AWS Batch: Monitoring and Load Testing

The series started with Wai Chee Yau’s exploration of AWS Batch and why we ended up choosing it…

Evaluating Apache Pulsar

Zendesk currently use Kafka, primarily for publishing of events. We are also considering publishing the state of entities to Kafka, using compacted topics to keep only the latest version of an entity. We do this already, but are considering it for much larger data sets. These compacted topics will be…

Datadog Log Management from Zero to One

Datadog APM Trace Search from Zero to One is the first part of advanced techniques we use at Zendesk to plan capacity and keep performance characteristics of the multi-tenant system at expected levels. Response times vary between different API endpoints and customers, as the data set grows rapidly, and new…