Nancy Yang

Building MaxHire, a AI-powered job application tracker that monitors job application emails through an event-driven pipeline and surfaces stats on a dashboard. Responsible for full-stack architecture, distributed backend services in Go, deployed on GCP Cloud Run. Focused on scalable data pipelines and reliable system design.

Led the network probing initiative from proof of concept to production — built an orchestration server that maintains a DC compute resource database via external APIs and streaming services, distributes probing plans to agents across tens of thousands of hypervisors, and provisions probing paths across DC network hierarchy layers. Emitted metrics established network availability SLOs and powered alerting for failed links.

Built an authentication gateway giving users a unified identity across multiple company websites, including a TypeScript package for authentication workflows and session management integrated with Ory/Kratos. Also developed a .NET hosted service that subscribes to Azure Active Directory notifications, syncing user profile updates and mapping AD roles to the internal permissions system.

Built and maintained several full-stack network management and video conferencing tools: Trac, a change request system for network operators with BGP statistics collection and automated router configuration to mitigate outages; a UDP probing scheduler with InfluxDB storage, Grafana dashboards, and a Google Chat bot for automated outage diagnosis; Uptime.vc, a scheduling platform for video conference calls with quality analysis and alerting; and Play.vc, a video call application with real-time call status via WebSocket and quality statistics.

Automated testing across several video infrastructure components: built Python modules to validate a patented SSD caching algorithm for Video on Demand servers based on content popularity models; developed tools to capture and analyze MPEG streams to verify packet repair on Video Quality Enhancement servers; simulated RTSP signals to test client kernel modules on IPTV set-top boxes, catching defects including a kernel memory leak and undelivered repaired packets; and stress tested the STUN module using custom Pagent packet templates.