> Amazon
Augmented reality across every product category
On the Visual Search and AR team, bringing immersive product experiences to the Amazon shopping app — mobile AR for TVs, furniture, eyewear, and cosmetics, each tailored to the product type and the shopping context around it.
01
TVs — a scalable architecture for View In Your Room
Rearchitected View In Your Room for TVs in the Amazon mobile app to support high-fidelity rendering and clearer usability. A custom rendering pipeline powered by Filament replaced the previous renderer, a new onboarding flow guided shoppers through the AR process, and the resulting foundation was extended to every category that followed.
- Modular architecture designed for cross-category expansion
- Real-time rendering via Filament integration
- Onboarding flow to orient first-time AR users
- Codebase reused across four-plus product verticals
02
Furniture — real-time customization and model swapping
Launched an AR furniture experience that let shoppers change product colors, swap between similar items, and place several pieces in the same scene at once. Multi-item placement made it possible to visualize a full room setup rather than one product in isolation, with interactive 3D cues guiding people through the interaction.
- Color and model swapping in real time
- Custom 3D instructional overlays
- Multi-item placement for whole-room visualization
03
Eyewear — face-aware virtual try-on
Partnered with brands including Ray-Ban to expand Amazon's 3D eyewear catalog and build a face-aware try-on experience. A real-time face detection pipeline placed frames accurately on the shopper's face, turning try-on into a selfie-based flow rather than a staged AR scene.
- Real-time face detection for frame placement
- Scaled 3D asset coverage across major eyewear brands
- Selfie-native try-on experience
04
Lipstick — ML-driven beauty try-on
Worked with Amazon's Graphics and ML research teams on custom models that render lipstick realistically across lighting conditions and skin tones. A static mode let shoppers test shades on preset faces spanning a range of skin tones, broadening representation for anyone not ready to point a camera at themselves.
- ML-based rendering for realistic lipstick application
- Inclusive static mode with multi-tone reference faces
- Mobile performance and latency tuning
Victor Liang's professional work, which the studio draws on. BlizL Labs is a separate, personal venture.