APT 304 | Our Work

LAVA : Google Cloud Migration

LAVA’s customers run in single-tenant environments. New customers meant lengthy manual provisioning and configuration of Nomad services. Cluster stability required frequent manual intervention, especially during node failures.

What we built

Customer environments were migrated from Nomad to GKE. Helm charts templated service configurations. Kubernetes service discovery simplified inter-service communication, and web services were exposed externally using Google Cloud Network Load Balancers. All services were automatically configured to report metrics and logs to a centralized dashboard.

Impact

The entire customer workload was migrated to GKE. New environments are now provisioned in minutes when a customer used to wait weeks. Autoscaling services reduced runtime cost by 23% per environment. Developer efficiency soared as service configuration and management time was redirected to product enhancement and customer engagement.

Read More Read Less

GoodTunes : Music Platform

Artists didn’t have an easy way to measure demand before printing limited run vinyl and collectibles. GoodTunes wanted to offer an all-in-one platform to connect with fans, monetize streaming, and determine vinyl demand with minimal risk for the musician.

What we built

We built a responsive web app and mobile apps for consumers to buy and listen to tunes, and an artist portal to manage launches and prepare vinyl pressings with payments facilitated through Stripe.

Impact

Artists launched albums with confidence, knowing they could size their vinyl pressing to demand. Artists outearn streaming services by up to 8x. Fans receive GoodDeed® Certificates for proof of purchase.

Read More Read Less

Paper Culture : Autoscaling GKE

Paper Culture’s e-commerce applications have a short, CPU-intensive period at startup, and afterwards CPU usage drops dramatically. Paper Culture’s engineers noticed that their GKE cluster’s node pools were almost always at maximum capacity, even when their applications had low activity.

What we built

We noticed that the Horizontal Pod Autoscalers was too aggressive. Configuring a waiting period after a scale-out event ensured that the CPU-intensive starting period had less of an impact on CPU usage averages. Resource requests and limits were tweaked to better reflect average values during the application’s lifetime.

Impact

The average node count in the Kubernetes clusters reduced by 75%. During peak usage, more nodes were created automatically. Outside of peak times, node pools were at minimum values. Average pod lifetime in Kubernetes increased to be closer to Paper Culture’s deployment cycle, with an 83x increase to 3.5 days.

Read More Read Less

Da Vinci : Handheld Warehouse Scanner Design

Warehouse teams are often forced to work with multiple poorly connected devices and applications to accomplish every day tasks, adding overhead as team members switch between tools. Da Vinci Unified tackled the problem head on.

What we built

We designed the RF mobile app specifically for in-warehouse conditions. Step-by-step guided workflows and scan validation prevents errors, streamlining processes on the floor. The same app drives receiving, putaway, picking, packing, replenishment, and shipping, reducing employee training and customer onboarding time.

Impact

Warehouses streamlined operations, leading to more successful workflow completions and fewer supervisor error corrections.

Read More Read Less

Durable Banking Sweep Network

A financial institution’s sweep network integration was flaky due to external delays and outages, often needing the engineering team’s intervention.

What we built

We used Temporal to incrementally stabilize the integration, moving steps into the durable workflow one at a time. Error prone steps automatically retried, and the workflow completed following the external system’s recovery. Asynchronous workflow design made it easy to automate the finance team’s approval each day, taking the engineering team out of the loop entirely.

Impact

The sweep integration ran without manual intervention. Incrementally pulling steps into the workflow minimized risk while targeting the least stable steps first. See our blog post for a detailed writeup.

Read More Read Less

Chatterworks : Recruiting Platform

Recruiters are bottlenecked by inefficient candidate sourcing and low outreach response rates. Chatterworks wanted to streamline high volume recruiting to help match great candidates with their ideal jobs.

What we built

We built an application to help recruiters find, reach, and pipeline hard-to-hire talent. Multiple data sources are merged to build complete candidate profiles. A browser extension surfaces comprehensive candidate data as recruiters search social sites. Candidates are automatically matched to open searches, with built-in multi-channel outreach to route messages where a candidate spends their time.

Impact

Recruiters find ideal matches for their open jobs, and save time by removing the need to manually aggregate profiles across social sites.

Read More Read Less

Food Truck App and OMS

Food trucks at festivals and events have huge lines, keeping attendees away from the fun. Trucks are congested as they take orders, handling cash or waiting for payments to process.

What we built

We designed and built web and mobile apps optimized for lightning fast customer ordering, an admin portal to manage menus and Stripe payments, and a mobile handheld POS to print incoming orders.

Impact

Ordering ahead, coupled with realtime notifications, reduced congestion around food trucks and stalls. Customers browse menus on the go and pay with their mobile wallet. Trucks focus on what they do best: making delicious food.

Read More Read Less

Ready to talk?

Book a free consult and let's talk through what you're building.