strengths.txt
Designs AI systems, not just AI features
▸
Architected an intent classification framework (5 intents × 3 contextual lenses × 7 user signals) that dynamically assembles search results for 3B+ users — a composable, systems-level approach that mirrors how LLM products need to be designed: adaptive, signal-aware, built for variance.
▸
Built LLM judge prompts for automated search quality evaluation across 3 workstreams → 16% improvement in issue detection, replacing manual review and demonstrating hands-on fluency with how language models fail.
Operates at the intersection of AI strategy and execution
▸
Contributed to CEO-level deck on AI assistant entry points; authored Search's AI mode strategy (entry points, pathing, cross-surface discovery) — reviewed at VP level.
▸
Led AI 2.0 launch design including full implementation plan and UI variants specifically engineered to mitigate engagement regressions — rare ability to think about AI rollout risk, not just AI potential.
Builds infrastructure others scale from
▸
Created an autonomous AI agent running 24/7 that detects UI bugs, generates code diffs, and conducts internal research — 48+ findings, now scaling org-wide. A principal-level move: solve your team's problem by building a tool the whole org adopts.
▸
Designed a unified component library averaging 3,200 insertions/week across 11 teams; intent taxonomy adopted org-wide.
Drives measurable outcomes at principal scope
▸
+75% global pivot volume · +9.65% content search DAP · +5.9% global search success · +4.6% ad score · 2× private shares — consistent pattern of shipping and moving numbers, not just shipping.
▸
Redesigned a core content discovery experience that performed so well the CEO debuted it publicly; became an app-wide priority for 2025.
Earns trust at the top, builds it across orgs
▸
Secured VP roadmap buy-in by connecting design decisions to company-level priorities. Elevated a P2 initiative to P0/P1 through a cross-functional brief — changed the roadmap through written clarity, not just influence.
▸
V-team design lead across 5+ designers and ~125 projects, aligning 5+ PMs, 4+ eng teams, research, and content design. Team exceeded DAU goals: 6.23% vs. 4.8% target.
Mentors toward promotion, not just performance
▸
Mentored IC5 → IC6 (promoted); developed another IC5 to "greatly exceeds" rating — a consistent pattern that signals genuine investment in others' growth. Table stakes for IC7.