Jason Gedge

I joined Gadget as the first engineer. I was involved in just about everything. Some of the major projects I worked on:

• Created an automated, zero-downtime migration for moving customers between database storage strategies. Migrated hundreds of thousands of tables with no downtime or data loss.
• Took our background actions project from 50% to 100%, with a focus on correctness — such as introducing idempotency into our retry logic — and the infrastructure to support it (new Temporal cluster).
• Initial implementation of @gadgetinc/mobx-quick-tree. Significantly reduced node event loop blocks, along with making GraphQL generation roughly 2x faster.
• Added user-defined HTTP routes for Gadget apps (e.g., similar to AWS Lambda or Google Cloud Functions, but slightly less general).

Non-technical things I was involved with:

• Sharing my knowledge and context of the system to help accelerate all members of our team.
• Work with other senior engineers to help unblock them and provide guidance on various projects.
• Informal mentoring via pair programming, code review, and internal conversations.
• Interviewing candidates to assess technical skill (and a small amount of sourcing).

• Helped build and influence the design of packwerk, an open-source tool for enforcing boundaries in Ruby/Rails codebases.
• Designed and developed an intra- and inter-cluster eBPF-based ping mesh for network observability in our Kubernetes platform. Required digging into the TCP stack implementation of linux (bootlin helped a lot for easy source code navigation). We sometimes detected networking issues in Google's datacenters before even they were aware of it.
• Built a data pipeline to isolate performance regressions in the Shopify platform. This was done in a Python notebook and could show both short-term and long-term regressions across many dimensions.
• Continued to help build our organization with regular interviews (~1 per week).

• Designed and developed Fraud Protect for Shopify Payments (now called "Shopify Protect for Shop Pay").
• Designed and developed a new model for expressing order risk for all of Shopify's orders, known as "risk assessments". Although it was used internally immediately, it wasn't exposed in the API until 2024.
• Improved Shopify's risk service by introducing a new pending state, which eased pressure on the service since the core Shopify service would often retry the same request (which was CPU intensive).
• Started interviewing on a regular basis for the R&D organization, primarily pair programming.

• Designed an approach for asynchronous label purchasing.
• Created a regional failover setup for our shipping service (using AWS Route 53).
• Automated our shipping service's infrastructure with Terraform.
• Moved our shipping service from classic EC2 to VPC EC2 with zero downtime.
• Improved performance of our shipping rates/label service by ~30% by through various small code changes discovered via profiling.

• Created new and expanded existing test infrastructure (e.g, static analysis, automated HTML validation, creating and validating test seams for functions with complex responses).
• Move the local development environments to the cloud, with a service that had pre-warmed environments and leasing capabilities. Although this significantly reduced boot time from several minutes to seconds, developers weren't as excited about the remote workflow.
• Improved the stack by dependency refactoring, reducing build times, build size, and resource consumption.
• Expanded YouTube's primary local development environment with new functionality, also building out new service fakes to improve boot time.