Hayden – Golang, Kubernetes, AWS

Self-storage operators need to control and monitor access to their facilities and individual units remotely, without staff on site. The platform connects gate keypads, door keypads, and smart unit locks over a mesh network to a cloud backend on AWS. Operators use it to manage tenant access, track facility activity, and forecast tenant behavior such as upcoming move-outs.

• Owned the tenant move-out prediction pipeline end-to-end: built the training dataset, trained a decision-tree model, and ran weekly inference on Redshift ML and SageMaker triggered by EventBridge, giving operators a weekly forecast of upcoming move-outs.
• Built the Go backend and the reporting frontend that surface move-out predictions to operators.
• Designed reliable async job processing with idempotency, database locks for short jobs, and expiring leases for long-running ones, plus health checks that kill stuck workers.
• Hardened the security of the platform with database encryption and cloud-native best practices.
• Led company-wide adoption of AI coding tools.
• Onboarded an overseas engineer and ran daily asynchronous handoffs with him.

Writing product descriptions by hand is slow for online store owners, especially with large catalogs. The app, built for Shopify merchants, generates product descriptions automatically from product images using AI image captioning. Merchants can fill out listings with ready-made copy instead of writing every description themselves.

• Integrated the AI image-captioning API so the app generates product descriptions directly from merchants’ product photos.
• Designed an asynchronous messaging system so description generation scales with catalog size without blocking the app.
• Built the Shopify app on Remix, TypeScript and React with a MySQL data layer, deployed on Kubernetes.

Employers increasingly offer student loan repayment as a benefit, but running the payments and bookkeeping by hand is slow and error-prone. The platform automates loan repayment processing and bookkeeping for employers. Employers and employees get dashboards to track repayment progress and benefit utilization.

• Built the employer portal in React with Chart.js dashboards showing benefit utilization and employee engagement.
• Built an internal admin console on React Admin with custom data providers for managing user accounts, loan configurations, and payment-processing workflows.
• Deployed the containerized services on AWS EKS behind an Application Load Balancer.
• Provisioned the AWS infrastructure with Terraform, including VPC, RDS PostgreSQL, S3 document storage, and CloudWatch monitoring and logging.

Logistics companies run large vehicle fleets where unexpected battery failures take vehicles off the road and fuel is one of the biggest operating costs. The platform collects telematics data from fleet vehicles and gives fleet managers predictive maintenance insights, fuel-usage analytics, and billing tools in one product. It processes millions of vehicle sensor records.

• Built machine learning models in PySpark on Databricks that predict vehicle battery failures from telematics data, including feature engineering and model tuning.
• Developed fuel-usage optimization algorithms with pandas, NumPy and SciPy based on historical driving patterns and vehicle performance metrics.
• Built customer-facing billing features in TypeScript and React, plus GraphQL endpoints in TypeScript and Go for billing data.
• Implemented internal microservice communication over gRPC and Protocol Buffers in Go.
• Provisioned AWS infrastructure with Terraform (ECS, RDS MySQL, S3 for invoices) and built Buildkite CI/CD pipelines with automated deploys to staging and production.
• Set up CloudWatch and Datadog monitoring with custom dashboards for billing-system health and performance.

A search evaluation system for the e-commerce search of a large home-improvement retailer. Without it, the search team could not reliably tell whether customers were finding the products they searched for. The system collects user interaction data from the search platform, turns it into search-quality metrics, and supports offline evaluation of search results against hand-annotated relevance data.

• Built the pipelines that collect user interaction data from the search platform and store it as search-quality metrics.
• Implemented offline search evaluation against hand-annotated relevance data, giving the search team a way to measure result quality before changes went live.