Joining Meta: Day One in Reality Labs
First impressions joining Meta's Reality Labs to lead Visual Positioning Service - different scale, familiar problems.
First week at Meta. Everything is bigger.
Scale Shock
Magic Leap was ~2,000 employees, with under 100 of us in perception. Meta is ~50,000+ employees, with thousands in Reality Labs alone. The numbers are one thing; the implications are another. More specialists, to the point that someone owns just the depth filtering. More process - reviews, approvals, documentation. More resources in compute, headcount, and budget, and with all of that, more complexity in coordination, alignment, and communication.
The VPS Mission
I'm here to build and lead the Visual Positioning Service (VPS): enabling XR devices to understand where they are in the world. The vision is that any XR device can localize in any mapped space, with 6DoF tracking relative to persistent world maps (the prerequisite I argued for in 2017), centimeter-level accuracy outdoors, and support across Quest headsets and future devices. It extends everything I did at Magic Leap, cross-device map sharing included, to global scale.
Different Challenges
The data alone changes the problem. Meta has billions of geolocated photos from apps like Instagram and Facebook, which could bootstrap worldwide mapping. Deployment is different too: Quest has millions of users, and features ship to everyone at once. And the infrastructure is Google-class - storage, compute, ML training. At Magic Leap we built custom everything. At Meta, we build on massive infrastructure.
Which problems stay the same?
Despite the scale difference, the core questions haven't moved. How do you build accurate 3D maps efficiently? How do you localize quickly on constrained hardware? How do you handle changing environments? How do you protect user privacy? The physics doesn't change because the company is bigger. I said when I announced the move that this was the same underlying problem at a different scale. One week in, that holds.
First Impressions
Talent density is high - every meeting has people who are world experts in something. Documentation is extensive; everything has a wiki page, a design doc, or a review thread. Process is real, and shipping anything requires multiple approval gates. And the ambition is extreme. The roadmaps assume success at everything.
Learning Mode
The plan for the first 90 days is to listen, learn, and build relationships: understand the existing systems, meet stakeholders across the org, figure out where I can add value, and build credibility through small wins. The temptation to immediately redesign everything is strong. I'm resisting it.
What I Bring
The relevant experience: end-to-end perception system design, synthetic data for perception training, hardware-software co-design, and shipping under constraints. What I need to learn: Meta's codebase and tools, the organizational dynamics, how decisions actually get made here, and what's already been tried.
Excited and humbled at the same time. The learning curve is steep again, which is exactly what I signed up for.