One Year at Meta: VPS Progress Report
Reflecting on a year of building Visual Positioning Service - shipped product, scaled team, lessons learned.
18 months since joining Meta, 12 since focusing fully on VPS. Time to assess.
What We Shipped
VPS Developer Preview:
- Live in 15 countries
- 500+ mapped locations
- Used by 50+ developer partners
- Integrated in Quest experiences
Technical Achievements:
- Sub-meter outdoor localization
- under 1 second query latency
- Privacy-preserving architecture
- Learned feature pipeline
Scale:
- Processing 10M+ images per day
- Serving 1M+ localization queries per day
- 50-person team
What Went Well
Team building: Hired strong people, promoted good managers, maintained culture.
Shipping velocity: Launched MVP in 6 months, iterated quarterly.
Cross-functional partnership: Worked well with Quest platform, privacy, policy.
Technical foundation: Architecture choices enabling scale and iteration.
What Didn't Go Well
Indoor localization: Behind plan. Harder than expected.
Mapping efficiency: Still too expensive per location. Economics don't scale.
Developer experience: SDK needs polish. Onboarding is confusing.
Accuracy consistency: Works great in some conditions, poorly in others.
Lessons Learned
Privacy is a design constraint, not a feature: Every decision must consider privacy from the start.
Coverage beats accuracy: Developers want it to work everywhere more than perfectly anywhere.
Operations at scale is different: What works for 10 locations doesn't work for 500.
Meta's resources are real: Infrastructure, talent, and reach enable things impossible elsewhere.
Looking to 2022
Priorities:
- Indoor localization launch
- Mapping cost reduction (10x)
- Developer self-service tools
- Quest 3 integration (next hardware)
Team growth: 50 → 80 (focused growth, not exponential)
Technical bets:
- Neural rendering for map compression
- Multi-modal fusion (WiFi + visual)
- Crowdsourced mapping at scale
Personal Reflection
What energizes me: Building systems that ship to millions, solving hard technical problems, growing talent.
What drains me: Meeting overhead, cross-org alignment, politics.
Growth edge: Delegating more, thinking bigger, influencing without authority.
Overall Assessment
VPS is real and growing. The path to ubiquitous spatial computing is clearer than a year ago.
We're not there yet - gaps in indoor, rural areas, edge cases. But the foundation is solid.
Excited for year two.