Portfolio

Stryker Buffington

Creative Technologist
AI Prototyper · Gameplay Engineer

I'm a software engineer and game designer who builds experimental interactive experiences. My work spans AI research, immersive technology, real-time game systems, and creative tools.

Currently AI Prototyper on AR glasses at Meta Reality Labs Research

01 — Featured project

NeoPoi Lightshow Editor

A desktop app for building music-synchronized lightshows for LED poi. I designed and built it with an AI-assisted workflow, and I use it to make the shows I perform.

  • Creative tools
  • Audio analysis
  • Generative AI
  • Pixel art
  • Hardware
The editor during playback. Left: both poi previewed as spinning discs. Right: the song map. Bottom: the waveform, song sections, and an image lane for each poi.
An exported show running on the poi.

The problem

LED poi don't have a screen. Each one has a single strip of LEDs that shows a narrow slice of an image at a time. Spin it, and your eye stitches those slices into a full picture. That makes the art awkward to author: images are tiny, they have to loop cleanly around a circle, and they need to change in time with the music.

Doing that by hand for a whole song takes a long time. I wanted a tool that would do the tedious parts, like finding the beats and sections and planning the changes, and leave me the creative decisions.

1

Song map

Local analysis on my machine finds tempo and beat candidates, silences, change points, and stutters. Optionally, the audio goes to Gemini, which labels sections (intro, verse, chorus…) with energy levels and short descriptions. Every section can be edited, split, merged, or dragged on the timeline.

2

Show direction

I choose how busy the show should be, for example a new image on every beat with strobes on stutters. The editor then turns the song map into a plan of image changes across both poi.

3

Images

Looping pixel-art images sit in a lane for each poi. The live stage previews them as spinning discs, in sync with the audio, so I can judge a section before it ever reaches the hardware.

4

Device & export

The finished sequence is exported to the physical poi, where it plays in time with the music.

How an image becomes a lightshow

The poi only ever lights one column. Each column is drawn at the angle the poi has reached, so the image wraps into a disc: columns become angles, rows become distance from your hand. Slow it down to see the slices.

Source image col 1 / 48

← start of loopend of loop →

On the poi

Illustrative only: a 48 × 14 pattern drawn in the browser, not NeoPoi's real resolution or output.

Why I care about this one

It's the most complete thing I've made on my own. An unusual idea went through tool and UI design, AI-assisted programming, audio analysis, and a hardware constraint, and came out as shows I actually perform. I made every product decision, and the tool kept changing as I used it.

02 — Experience

Where I've worked

Most of my career has been spent next to researchers, turning new sensing and ML capabilities into things people can pick up and use. One stretch was spent shipping a production game system.

  1. Jul 2025 — Present

    AI Prototyper, AR Glasses

    Meta · Reality Labs Research

    AI-first interactions for AR glasses, built on input technology that's still in research.

    • Prototype AR-glasses experiences in Unity using eye tracking, hand tracking, and EMG gesture input, and present AI-first UX concepts to research leadership.
    • Built the RESTful networking behind proactive-agent prototypes that infer intent from multimodal signals: gaze, hand interactions, voice events, and scene understanding.
    • Partnered with external researchers to apply computer-vision algorithms to vehicle-mounted sensors for context-aware features.
    • Ship exploratory demos for stakeholder reviews and for tech transfer to product teams.
    • Unity
    • C#
    • Multimodal AI
    • Eye tracking
    • EMG
  2. Mar 2024 — Apr 2025

    Mixed Reality Software Engineer

    Microsoft · HoloLens

    Shipped the Squad Immersive Virtual Trainer, a customer-facing Unity mixed-reality training system for the US Army.

    • Designed, documented, and implemented gameplay, game AI, and user-generated content systems in six-week cycles.
    • Built real-time AR features involving multiplayer networking, world-locked holograms, tight performance budgets, and device setup and rostering.
    • Ran LiveOps for a deployed product: code review, bug triage, and fast fixes.
    • Worked across engineering, QA, animation, and art.
    • Unity
    • C#
    • Gameplay
    • Game AI
    • Multiplayer
  3. Jan 2022 — Jan 2024

    VR Prototyper

    Meta · Reality Labs Research

    VR, haptics, and machine-learning prototypes built alongside research scientists.

    • Designed and built the VR demo that was the first successful demonstration of a novel haptic technology to company stakeholders.
    • Engineered a hand-tracking gesture system using Gaussian-process modeling, computer-vision hand tracking, and linear algebra, which unblocked a team of haptics researchers.
    • Shipped an open-source Unity/C# API for AEPsych, Meta's adaptive-experimentation ML tool. It covered the client/server protocol, docs, and support, plus eight interactive demos and a conference presentation.
    • Built four more proof-of-concept VR experiences and data-collection environments with a team of 10+ researchers, designers, and engineers.
    • Unity
    • VR
    • Haptics
    • ML
    • Research prototyping
  4. Aug — Sep 2021

    Machine Learning Research Intern

    AIST · Tokyo, Japan

    Simulated shoppers that helped evaluate store layouts for social-distancing compliance.

    • Trained shopper agents with Unity ML-Agents using reinforcement learning and imitation learning, then compared the two approaches.
    • Built a multiplayer VR shopping simulator to record human demonstrations for the imitation-learning agent.
    • Collaborated with AI researchers across time zones.
    • Unity ML-Agents
    • RL
    • Imitation learning
    • VR
Earlier
  • Multisensory VR physical-therapy research, ASSIST Labs, UC Santa Cruz · 2019–2022. Led a team of three on a stress-reactive VR horror game with a GSR and heart-rate pipeline into Unity, and was lead author on the paper presented at IEEE SeGAH.
  • Co-founder, GetVirtual · 2020–2021. A nonprofit pairing student teams with local businesses on web and digital work.
  • Unreal Engine instructor, iD Tech at Stanford · Summer 2019.
  • Collection digitization intern, Smithsonian National Museum of Natural History · Summer 2018.

Education

University of California, Santa Cruz

B.S. Computer Science: Game Design, with an emphasis in VR and AI

2018 — 2022 GPA 3.91

03 — Other work

Research, games, and prototypes

2022 · Meta Reality Labs

AEPsych VR Throw Optimizer

A VR throwing interaction that tunes itself. AEPsych's continuous-outcome models search the interaction's parameters for the settings that make each player throw most accurately. I proposed the demo and built the gameplay and Unity client.

  • Unity
  • C#
  • Adaptive ML
  • VR

2022–2023 · Open source

AEPsych Unity client

The Unity/C# front end and client/server protocol that let Unity developers run AEPsych experiments. I also wrote the documentation and handled support.

  • C#
  • API design
  • Python server
Internal research · no public media

Meta Reality Labs

Haptics VR demonstration

An immersive VR demo built to show a novel haptic technology. It became the technology's first successful demonstration to company stakeholders.

  • Unity
  • Haptics
  • VR
Internal research · no public media

Meta Reality Labs

Hand-tracking gesture system

Gesture recognition built on computer-vision hand tracking and Gaussian-process modeling, written to unblock a haptics research team.

  • Computer vision
  • Gaussian processes
  • Unity

2021 · AIST, Tokyo

ML Shopping Simulator

Shopper agents trained in Unity ML-Agents to test how store layouts affect social distancing. Human play in a multiplayer VR store provided demonstrations for a GAIL agent. A pure reinforcement-learning agent ended up producing the more usable simulation.

  • Unity ML-Agents
  • RL
  • GAIL
  • VR

2020 · UCSC Game AI

VR Tower Defense

A VR horror game that reads the player's skin response and heart rate, then uses a stress-controller state machine to change the game environment in response.

  • Unity
  • Biofeedback
  • Game AI
Re-Pair title screen Play in browser

2021 · Phaser

Re-Pair

A story-driven puzzle platformer about rewinding time to save someone you lost. I proposed it, produced it, and was the only programmer.

  • JavaScript
  • Phaser
  • Puzzle design
HenHop title screen Play in browser

2021 · Phaser

HenHop

An endless runner over procedurally generated terrain. Last long enough and it switches into Nightmare Mode. Team of three. I was the only programmer.

  • JavaScript
  • Procedural generation
Ringout Avalanche Layers

2020 · UCSC rapid prototyping

One-week arcade prototypes

Small games built in a week each with crisp-game-lib, each to a strict constraint. Ringout's brief, for example, was "one button." Click a screenshot to play.

  • JavaScript
  • Rapid prototyping
  • Game feel

04 — Skills

What I work with

Brighter items are the ones I use daily. The rest I've used for real projects, but less often.

Creative & interaction

  • Gameplay prototyping
  • Rapid prototyping
  • Interactive experience design
  • XR interaction design
  • Game design
  • Experimental interfaces
  • 0-to-1 product work

Engineering

  • Unity / C#
  • Real-time systems
  • Tool development
  • API design & integration
  • Python
  • JavaScript / React
  • C++
  • Networking
  • Git

AI & research

  • AI-assisted development
  • Multimodal AI interactions
  • Generative AI workflows
  • LLM integration
  • Unity ML-Agents
  • Reinforcement & imitation learning
  • Computer-vision integration
  • HCI research

Hardware & platforms

  • AR / VR / MR
  • Eye & hand tracking
  • EMG input
  • Haptics
  • Unreal Engine
  • PyTorch
  • WebGL

05 — Flow arts

Poi_Fish

Outside of work I spin poi: LED poi, fire poi, and silk. I choreograph sets to music and post them as Poi_Fish.

It's also where a lot of my design taste comes from. A good spin and a good game interaction have similar requirements: readable motion, timing that lands on the beat, and a short loop between trying something and seeing it. NeoPoi exists because I wanted better tools for my own art.

  • LED poi
  • Fire poi
  • Silk poi
  • Choreography
  • Music-synced visuals

@poi__fish on Instagram TikTok

Stryker Buffington

06 — About

About me

My favorite part of the work is first contact with a new capability: an eye tracker, an EMG band, a haptic actuator, a model that can describe a song. Figuring out what it's actually good for usually means building something quickly, putting it in someone's hands, and changing it until it feels right.

I studied computer science with a focus on game design at UC Santa Cruz. Since then I've worked as a research prototyper and as a production game engineer, and I use AI tools heavily in how I build. Lately I've been most interested in real-time generative models and the kinds of play they could make possible.

Away from screens: circus arts (aerial silks, trapeze, lyra), jiu-jitsu, tennis, and taking my cat Salem to places cats don't usually go.