Spark AR

Before Spark AR existed as a named product, the Messenger AR team shipped four flagship effects to prove AR could be commercial-grade across very different verticals: Sephora virtual makeup, KIA at vehicle scale, Nike SNKRS footwear, and ASUS Portal hardware visualization.

I presented them at F8 2018 Day 2, Class 113 — thrown to by Loredana Crisan, then Head of Design at Messenger (now CDO at Figma). The room was rated for 100 and packed past capacity. The talk made the case for AR being a primary social mode, not a filter add-on.

That session is the inflection point for everything below.

What grew out of Class 113 was Spark AR — runtime, capabilities, and creator tooling, supported by a 400+ person XFN organization. I led the Experiences team in close partnership with the London-based creator-tooling team, scaling from 2 to 22 ICs over five years.

The org introduced the Product Design Prototyper role within Spark AR, lifting prototype velocity by 60+%. We also designed and shipped Story Time with the Portal team — a flagship AR feature for video calls that outperformed standard video on engagement and retention. The result confirmed AR could be a primary social mode, not a filter add-on.

In parallel with platform growth, the team took AR Commerce from prototype to launch across Instagram and the Family of Apps. Try-ons (cosmetics, footwear, eyewear), branded AR ad formats (Pringles, the Avatar movie launch), and dynamic product placement — with 10+ retail partners and 200+ products in market.

Underneath sat the Container Effects System — an extensible substrate for portable UX across Instagram and Facebook, separating interface from content. Three pillars (Bulk Asset Upload · Parameterization · Content Assembly) and seven components made AR Commerce buildable at retailer scale rather than per-effect bespoke.

The work surfaced a Three-Pillar success model balancing consumer · platform · business outcomes, and a conversion funnel that became the template for measuring AR-driven commerce inside Meta.

The marketplace · the system · the experience

Spark AR was a developer platform. The work that scaled it past flagship demos was teaching the creator community how to think about AR. I co-wrote the IC public talk track at F8 (multi-year) and Oculus Connect 6, and partnered with UX research to publish the public-facing Spark AR Design Guidelines.

The core methodology was the AR Fundamentals — five UX Touch-Points and four North Star Beliefs distilled from running the Experiences team: hypotheses we’d proven, hypotheses we’d had to throw out, and the patterns that survived contact with creators and users.

The point was to teach, not prescribe.

ARBE was a research arc inside Spark AR exploring what AR could do beyond face filters and surface placement — specifically, location-based AR: world-anchored content, persistent objects across sessions, multi-user shared spaces, and the consequence of AR leaving the phone screen.

The work generated a public patent — world-locked interactive avatars — and the design vocabulary that fed into the Quest MR Mode and the Orion AR Glasses Day-1 use cases. ARBE is where Spark AR’s mobile language started to become a glasses language.

Mobile AR taught the surface. ARBE taught the space.

World AR Avatars · the patent in context

The most concrete instance of the patent was World AR Avatars — a research-to-product arc that put a personal Avatar into the world-camera so people could express themselves in space, not just on their face. The hypothesis: world-anchored AR is more popular when it includes people. The data hook: 30% retention for content with people in it vs 12% for non-people World AR. We launched the MVP in the FB camera in Cambodia first, then iterated on inventory and location behaviours.

Read the full deck · 26 pages, gated

F8 2019 was the moment Spark AR went from internal team to public platform. I gave the flagship outlook talk, and supported two team members on stage with their domain-specific deep-dives. Three angles on the same year of work.

Spark AR’s runtime, design language, and team became the foundation for the Orion AR Glasses Day-1 experiences, the Quest MR Mode, and the FAIR Embodied AI program (2023–2025) — where, as Senior Product Design Manager, I worked across multiple teams on AI agents for glasses, headsets, and robotics.