Steven Bryan

Engineered evaluation infrastructure for Google’s Computer Use Model launch, executing benchmarks against OpenAI and Anthropic models and producing performance data used in launch day marketing materials. Architected an end-to-end task runner using Stagehand and Braintrust to automate browser task execution while capturing comprehensive logs and screenshots across 5,000+ automated test runs. Developed the “Speech to Browser” (SToB) demo with Cartesia, leveraging low-latency voice inference to drive browser actions and showcasing it to strategic partners to validate voice-driven automation use cases. Contributed production code to Stagehand v3, improving API stability and documentation clarity for the open-source community. Scaled the developer community by 28% via technical advocacy on Discord/Slack, providing high-touch support to enterprise customers including engineers at Ramp and Vercel.

Delivered production React/Node.js web applications for clients, integrating complex PostgreSQL schemas for data-intensive features. Maintained a 99.9% uptime SLA on AWS infrastructure (EC2, S3, RDS) through load balancing and monitoring configuration. Designed scalable RESTful microservices using Express.js and Auth0, implementing secure role-based access control (RBAC) for client user bases. Optimized the CI/CD deployment workflow using GitHub Actions, reducing build-to-deploy time by 40% and eliminating manual deployment errors.

Engineered a custom Machine Learning training pipeline that cut experiment iteration cycles to roughly a quarter of what they were. Spearheaded technology transfer initiatives, collaborating with early-stage startups to convert research-grade ML prototypes into production-ready APIs.