Experience snapshot
Apple Maps
Human data & ML operations
Analyst, Product & Operations.
Built the human-data, quality, and operating systems behind cross-functional ML programs—turning ambiguous model failures into measurable improvements without slipping delivery.
The context.
Problem space
Model quality and 3D coverage depended on reliable labeling and QA across large distributed workflows while technical requirements continued to evolve.
Responsibility
Own delivery across ML engineering, operations, vendors, a 35-person analyst team, and a 60-person editing organization; translate quality issues into scoped iteration cycles.
Outcome
Improved model quality and operational reliability while sustaining throughput and meeting every committed milestone on schedule.
What I owned.
01
Built the KPI framework, weekly operating review, and delivery cadence for a six-month cross-functional automation program.
02
Managed a 35-person analyst team and coordinated a 60-person editing organization producing roughly 1,000 submissions per week; rebuilt labeling standards, QA sampling, and throughput targets.
03
Traced stalled 3D coverage back to source-data gaps and structured-output defects, scoped the fix with engineering, and drove it across 10+ major U.S. markets.
04
Converted ambiguous quality complaints into testable requirements and iteration cycles, contributing to 96% signage-detection accuracy across real-world road environments.
05
Built the reporting layer for throughput, defect rate, and delivery timelines and coordinated 3D model production and refinement across six global cities for Apple Maps Flyover.
The evidence.
97%
model accuracy after rebuilding labeling and QA systems
~1K / wk
human-data submissions sustained across the editing workflow
100%
committed milestones delivered on schedule
A note on confidentiality.
This experience involves internal and proprietary work. The overview focuses on my public-safe responsibilities, operating systems, and approved outcomes; confidential interfaces, datasets, implementation details, and internal company information are intentionally omitted.
PORTFOLIOBack to home
↗