• Own core components of Meta’s AI/ML training and evaluation data platform for LLM and multimodal workflows - including data qualification, routing/orchestration, quality control, and fallback mechanisms.

• Designed and scaled routing for 15M+ weekly labeling/review/eval tasks across 80+ markets, improving stability and throughput.

• Own finance tooling and data integrations supporting a multi-billion-dollar vendor ecosystem, ensuring accurate cost/signal attribution and reliable downstream behavior across dependent systems.

• Built evaluation workflow patterns that increase signal integrity and reduce quality variance across multi-stage review.

• Drive multi-year platform roadmap, OKRs, and analytics; translate ambiguous model/support needs into measurable requirements and sequenced delivery.

• Introduced experimentation and metrics guardrails to de-risk changes at scale and standardize launch/readout practices.

• Partner with Engineering, ML/DS, Research, and cross-functional stakeholders to align tradeoffs (quality, cost, throughput) and sustain operating rhythms.

• Mentor partners on instrumentation and KPI design, raising the bar for measurement and decision-making.

• Fixed routing and work-delivery defects on a critical workflow, improving SLA performance by 23% and stabilizing reviewer throughput without additional headcount.

• Implemented forecasting and capacity models (Prophet) for high-volume review workflows to improve staffing accuracy and planning cadence.

• Scaled time-tracking and compliance workflows across regions; secured SWE investment to replace legacy supply/demand planning tools.

• Built early global workforce supply optimization processes that later informed orchestration patterns used in large-scale review workflows.

• Reduced hate-speech workflow data reporting lag from 24h → 30m through process redesign and tooling improvements, enabling accelerated review cycles.

• Defined routing configuration and skilling frameworks across 80+ markets, improving match quality between work types and reviewer capabilities (i.e. supply vs demand).

• Built predictive staffing and cost-impact models linking operational volume, training throughput, and financial planning.

• Led rollout of forecasting and scheduling tooling across APAC markets; established reporting, adoption KPIs, and operating cadence.

• Led multi-market customer support teams; improved CSAT from 75% → 88% through coaching, peer-learning loops, and data-driven QA insights.

• Reduced repeat contacts and stabilized handle time through root-cause analysis and workflow improvements.