royston@prod:~$ whoami
role: Staff Software Engineer (IC6)
loc: San Francisco Bay Area
I build backend systems for large-scale ML and generative AI products — real-time inference under production latency, cost, and integrity constraints, evaluation infrastructure for generative models, and ML internationalization for low-resource markets. Fifteen-plus years across high-stakes domains: risk systems on Wall Street, and consumer AI (currently at Meta) serving hundreds of millions of people daily.
royston@prod:~$ cat experience.log
Technical lead for GenAI Suggested Replies, Contextual Replies, Notifications, and Feedback workstreams.
Counterparty-risk (CVA/RWA) and CCAR stress-testing infrastructure; grid-computing optimization.
royston@prod:~$ ls selected-work/ --featured
├── genai/suggested-replies
Backend lead — integrating large language and visual content-understanding models into a real-time consumer product. → shipped under production latency, cost, and integrity constraints at consumer scale.
├── eval/launch-gating
Structured human evaluation at scale, secondary-LLM prompt-quality pipelines, and per-suggestion engagement measurement. → the gate GenAI features pass before launch.
├── i18n/conversational-suggestions
Led integrity recovery and expansion of ML-driven suggestion systems to non-English markets. → translation-pathway architecture for low-resource deployment.
├── infra/autogen
Server-side interactive-overlay generation and the code-generation framework behind interactive story feedback formats.
└── shipped/
Backend lead for Candid Stories (dual-camera format, 2022); built the Stories Archive with its Activity Log and data-portability layer; shipped interactive polls, reactions, and sticker feedback.
royston@prod:~$ history | grep -i "service"
royston@prod:~$ ./reviewing-interests --list-flags