Brad Rathke

Sole platform engineer for a social platform serving 100k+ users through public beta. Built and operated full AWS infrastructure with Pulumi IaC — ECS Fargate, Aurora PostgreSQL, Neptune, CloudFront, and cross-cloud auth to GCP Vertex AI. Delivered zero-downtime deployments, autoscaling across every production service, and a complete observability stack with OpenTelemetry and Grafana Cloud. Increased infrastructure reliability from 97.19% to 99.98%. Owned CI/CD and developer experience end-to-end, from GitHub Actions pipelines to Tilt-based local dev environments.

Founding engineer. Architected the back-end platform from the ground up in Go, GCP, and Kubernetes, launching the core product in June 2025. Designed the flagship semantic search system powered by LLMs — the key technical differentiator at launch. Co-invented “Pivots,” a novel search refinement approach that drove significant increases in user task completion (U.S. patent application filed 2025). Built and led the initial backend engineering team, establishing coding standards and development culture.

Architected and led initiative to optimize the Stadia video pipeline, reducing encoder-to-GPU latency by 10x through DMA-BUF sharing. Contributed to Stadia’s Windows compatibility layer for low-modification application porting. Integrated Stadia’s low-latency streaming into Android XR, preserving core technology value post-shutdown. Prototyped real-time eye-tracking for dynamic foveated encoding in VR.

Developed Vulkan swapchain implementation enabling direct-to-encoder video frame transport with sub-perceptible encoding latency. Shipped platform-differentiating features from concept to launch, including StreamConnect (PIP teammate views) and Couch Co-Op via stream splicing. Led PCI-Express validation toolkit for next-generation server hardware. Tech Lead for visual perceptual quality metric infrastructure used by video and network teams to iterate with confidence.

Developed and optimized Metal and OpenGL graphics drivers for macOS, meeting Apple’s strict launch criteria. Built system to manage and merge driver code across multiple GPU generations, reducing feature drift. Performed pre-silicon validation for next-generation integrated GPU hardware. Deployed distributed build caching, slashing team-wide C++ build times by an order of magnitude.

Implemented sensor data fusion and track correlation algorithms for MH-60 Romeo and Sierra helicopter platforms, enabling real-time entity resolution for flight crews. Developed mission-critical software on VxWorks RTOS under stringent real-time constraints. Built Python data analysis tooling to evaluate algorithm performance. Conducted closed-laboratory verification in simulated helicopter cockpit environments with synthetic track scenarios. Modernized legacy development workflows, reducing development cycle times across the engineering organization.

Contributed to Intel’s OSPRay raytracing framework, helping establish Intel’s position in the high-performance raytracing market. Designed an OpenGL-compatible raytracer API that reduced customer migration barriers from traditional rasterization pipelines. Published peer-reviewed research at EGPGV 2015.

Led the company’s cross-platform transformation from Windows to Linux, directly enabling a major Boeing contract for GenesisIG image generation. Developed real-time rendering software in OpenGL and C++ powering flight simulators used for pilot training. Innovated a stochastic dithering algorithm that eliminated color banding on aviation displays, meeting strict certification requirements.

Thesis: The Visualization of Unstructured Polyhedral Grids

Dan Cary, Brad Rathke

Rathke, Wald, Chiu, Brownlee, pp. 33-41

Rathke, Brad