Projects

I led Rendering QA for Spatial Scenes on iOS, iPadOS, and visionOS, reporting on visual quality across 12 GPU families during feature development. Built automated rendering test suites, triaged complex failures, and performed targeted manual testing on edge cases. Additionally, drove QA efforts for the visionOS API enabling third-party developers to create spatial scenes using RealityKit's ImagePresentationComponent.

Introducing Spatial Scenes

ImagePresentationComponent

Lots I could write about here, but key highlights include:

• Sole 3D modeler for the life span of the company, owning all work related to 3D modeling of custom assets, handling of 3D model uploads from users, 3D model optimization pipelines, and enterprise 3D modeling requests.
• Led design and implementation of 30+ customizable 3D templates and 100s of 3D models from scratch, optimized for browser rendering with multiple UV channels to allow tileable texture manipulation while preserving light maps in real-time browser environment.
• Created production-ready AR/VR-compatible 3D assets for both free users and enterprise clients, often converting complex files (Unity, Unreal, FBX, OBJ, USDZ) into streamlined, web-optimized formats.
• Engineered solutions for responsive, real-time rendering across browsers and devices, maintaining seamless UX/UI interactions critical for high-profile client launches and virtual spaces targeting iPhone 6 Safari support.
• Relentlessly pushed tools on weekly iteration cycles, driving product decisions from daily user interviews over the course of 3 years.
• 5,000,000+ unique site visits; 150,000+ collective publishes; 50,000+ assets uploaded; 15,000+ unique sites created

Aside from adopting a performance first 3D modeling mindset while building and optimizing models, I contributed the following to further boost performance on user sites:

• Built automated and manual 3D model optimization pipelines, drastically reducing models from gigabytes to web-friendly sizes (often from 10GB to under 10MB), ensuring smooth performance on devices as limited as iPhone 6 hardware.
• Developed and fine-tuned GPU texture compression workflows (ASTC, UASTC, KTX), balancing visual fidelity and performance, critical for client launches attracting 10,000+ simultaneous users.
• Incorporated many WebGL rendering performance techniques into our pipelines leveraging mesh decimation, LOD, selective polygon culling, texture atlassing, on the fly collider generation, instancing, baking high-poly normals onto optimized low-poly models, etc.
• Collaborated with glTF Transform creator Don McCurdy to integrate palette texture support, significantly reducing draw calls for models with multiple meshes using the same material properties.

Source

Each of the above videos were created from scratch and completed start to finish with a 2 day turnaround window. The FBRC video was a teaser demo made while trying to land an enterprise client, and the collaborative editing launch was a marketing video announcement.

I love watching television... when I'm with good company. This prototype explores how it could feel watching entertainment alongside your favorite players and characters.

I manually rigged the 3D character models and triggered their animations + dialogues to play at set AVPlayer timestamps. I created the voice cloning with Eleven Labs and wrote the script myself.

Providing fresh, interactive content to old episodes feels like low hanging fruit inside Apple Vision Pro — a DLC for TV type experience (MST3000 style).

I think majority of people would find this concept overwhelming. But I believe there are times when advertisements can seamlessly integrate into experiences that insinuate feelings of delight. Interstitial ads with subtle incorporation of 3D content — executed thoughtfully — could be something that viewers might enjoy if done right.

This was built by using #EXT-X-DATERANGE tags fetched from .m3u8 files, opening and dismissing mixed content inside the space according to the playback of mid-roll time sets.

This was my first time working with the HLS spec, and it made me wonder "could this spec be ammended to allow for fetching of .usdz files?" I would be interested to see how content providers could leverage this to create engaging Super Bowl commercials, or offer opt-in 3D experiences for ad viewing. Even products or movie trailers could persist 3D elements from an advertisement as a volumetric window that lives in your space while you continue watching. Feels like a truly new way to ingest media.

Coming from a 3D modeling background, RealityKit's limited animation support—particularly its inability to handle .usdz files with multiple embedded animations—posed significant challenges. To overcome this, I developed a workflow involving individual .usdz animation files, extracting their animation data into structured arrays, and applying them across consistent skeletal models standardized via a custom Python script in Blender.

To simplify animation sequencing, I prototyped a drag-and-drop interface allowing dynamic ordering of animations, offering significant usability improvements over manual adjustments in Xcode. Building upon this, my goal is to create a tool enabling intuitive sequencing, timing control, and looping options, that could then be baked out into one animation sequence.

I think splash screens have the potential to feel really magical on this platform. These are some prototypes I made with the Portal Component inside RealityKit, leveraging particle emitters and text views to welcome a visitor on app launch.

I gather a lot of inspiration from the Suprematism and Lettrism movements. I also work on a zine with friends in San Francisco (currently working on our 5th issue). And I've been creating ceramics on and off for 12 years.

View more on Instagram