While our client’s business was rapidly signing up customers, the burden of developing and supporting an overly complex monolithic application was quickly becoming a bottleneck, both from an effort perspective and (cloud) infrastructure cost perspective.

Accordingly, they mobilized an effort to transition all functionality (over time) to a new microservices architecture and brought us in to work with key members of the development, operations, and QA teams to lead the design, delivery, and training on new Kubernetes-based infrastructure on AWS, built as-code with Terraform, offering CI/CD through Jenkins.

In one year, their new microservices were serving all production customer traffic (while still calling out to legacy systems on the backend when needed), being updated during business hours (without downtime) instead of nights and weekends, and autoscaling to minimize costs, all while operations ran automatically and transparently.

Our client was administering a Kubernetes-based platform on GCP (GKE) being used by numerous internal teams to run and manage their workloads.

To ensure platform stability ahead of go-live, they required comprehensive monitoring and efficient alerting on the platform's ability to run and manage workloads.

In six weeks, we collaboratively developed a list of KPIs and delivered automation (via Terraform) to export state and performance data from their platform, visualize it in dashboards built with GCP MQL, and alert the client's PagerDuty app upon violation of configured thresholds.

While our client’s business and capabilities were growing, so were their security needs around data loss prevention (DLP).

Accordingly, they brought us in to design and deliver a custom intrusion prevention system (IPS) built around Suricata, Python, and Docker to inspect all egress traffic from their cloud VPCs and block traffic upon detection of information matching any configured patterns (defined as regular expressions) as well as alert security response teams as appropriate.

In six weeks, the system was running in our client’s development VPCs, and in another six weeks in production VPCs, as well as enjoying automated deployment/updates via integration with their existing Puppet-based infrastructure.

Our client was administering a Kubernetes-based ML platform on GCP (GKE) being used by numerous internal teams to train and serve models.

As adoption grew, so did a glaring need for cost attribution on resource usage by each team, including CPU, memory, storage, and particularly GPU given its increased demand (globally).

In four weeks, we delivered automation (via Terraform) to export resource usage data from their Kubernetes clusters and a comprehensive SQL query to calculate costs and group by team by correlating to client-internal tables for resource pricing and team membership.

Our client was rolling out a new line of smart speakers which could connect to APIs running in their AWS VPCs to request information. As part of this effort, they needed a scalable cloud-based API development platform that could maximize productivity for hundreds of developers while minimizing operational costs.

We were brought in to lead the design and development of the application configuration, CI/CD, and monitoring aspects of the platform, which was built with Kubernetes/Docker, Nginx, Jenkins, Prometheus/Grafana, Terraform, and a custom CLI written in Python to give developers the ability to leverage all of it.

In six months, their initial set of APIs were in production. Today they support capabilities for millions of devices in the client’s ecosystem of smart IoT products.

Our client was looking to automate performance testing and reporting of their new microservices.

We were brought in to containerize their Locust test scripts and create a series of Jenkins jobs that would automate their deployment on Kubernetes.

In four weeks, developers were able to use a single Jenkins job to run performance tests for any microservice(s), including the ability to set parameters (e.g. requests/second) and define thresholds for success (e.g. 98th % latency) as well as track historical results of all performance test runs.

Our client was developing a mobile app to help skiers collect and analyze their performance data in real-time. They partnered with a former Olympic gold medal skier to develop the universal alpine ranking (UAR) system to calculate an overall rating for a skier (between 0 and 100) based on how they could perform against a series of challenges that analyzed various data collected from sensors on their skis and body, e.g. velocities, G-force, hip angulation, and airtime.

We were brought in (with minimal knowledge in skiing) to work with the Head of Product and the Chief Scientist to deliver C++ code that could be run within their mobile app to analyze incoming sensor data and calculate UAR scores in real-time.

In six weeks, skiers were using the app to measure their UAR scores while skiing local mountains.