# Technical Skills Matrix
Specific skills for each role in [[Roles and Hiring Plan]]. Use these lists for job descriptions and interview checklists. Training and tutorials come next, in each folder's *Tutorials* note.
## Character art
Stage 2 in depth (sculpt, Mesh to MetaHuman, Creator, MetaHuman for Maya, grooms, cloth): [[Stage 2 — Likeness and Body Skills]]
**Character artist (MetaHuman / Maya)**
- MetaHuman Creator in UE 5.8: parametric body, face blending, skin, teeth, eyes, custom lighting preview
- **Mesh to MetaHuman** from sculpts or scans (head and body, arbitrary topology)
- Sculpting in ZBrush or Maya; retopology; UV layout
- Likeness from photographs and portraits; photogrammetry cleanup (RealityScan or similar)
- Skin look-dev: unbaked textures and material control in MetaHuman 5.8
- LOD authoring: Cine vs Optimized assemblies, texture budgets for Quest and phones
- Maya ↔ Unreal round trip (FBX, USD); MetaHuman for Maya
**Groom artist**
- UE Groom (Alembic) authoring; XGen or Houdini hair workflows
- Strand hair for kiosks; **hair cards** for Quest, Vive and phones
- Period hairstyles and facial hair from references
**Costume designer / cloth artist**
- Period costume research
- Marvelous Designer or CLO patterning; simulation; baking to MetaHuman Outfits so garments refit across bodies
- Fabric materials (wool, linen, silk) at realistic scale
**Environment artist**
- Architectural reconstruction from HABS drawings and site photos
- Nanite, Lumen / Lumen Lite, MegaLights (production-ready in 5.8) for kiosk; baked lighting for Quest
- Modular kits reused across rooms; draw-call and texture budgets for standalone headsets
## Animation and performance
**Technical animator / rigger**
- MetaHuman Animator: iPhone/TrueDepth, stereo head-mounted camera, webcam/mono video, **single-camera body capture (Windows)**
- Live Link Face (iOS and Android), RTSP streaming input
- MetaHuman DNA and RigLogic (open source under MIT since 5.8); Control Rig, Direct Mesh Controls, Control Rig Physics
- Retargeting (IK Retargeter), animation blueprints, blend spaces for idles and gestures
- Maya: mocap cleanup, HumanIK, skin weights (ngSkinTools in Maya 2027), FBX/USD export
- Baking and compressing animation for mobile (USDZ/glTF skeletal animation)
**Mocap technician**
- Inertial suits (e.g. Rokoko, Xsens) or optical stages (Vicon, OptiTrack)
- Calibration, solving, data cleanup; syncing face, body and audio timecode
## Engine and platforms
**Unreal engineer**
- UE 5.8 C++ and Blueprint; Sequencer; state machines for the program clock
- Audio: MetaSounds, submixes, captions and subtitles
- Pixel Streaming 2 and its signalling/front-end infrastructure
- Packaging for Windows; crash reporting; kiosk auto-start and watchdogs
- Source control (Perforce or Git LFS); build automation (Unreal Automation Tool / BuildGraph)
**XR engineer**
- OpenXR in Unreal; Meta XR plugin; Android XR plugin; Vive OpenXR
- Performance on mobile GPUs: forward shading, fixed foveation, Application SpaceWarp, texture compression (ASTC)
- Comfort design: seated mode, teleport, vignettes
- Fleet: ArborXR or ManageXR enrollment, kiosk/shared mode, remote app updates
**visionOS developer** (only if in scope)
- Swift, SwiftUI, RealityKit, Reality Composer Pro, USD
- Or Unity 6 + PolySpatial (URP, Shader Graph only); full-immersion Unreal builds
- Xcode on Apple silicon; TestFlight; enterprise distribution
**Mobile AR engineer**
- ARKit + RealityKit (iOS); ARCore (Android); or Unity AR Foundation for both
- USDZ (Quick Look) and glTF (Scene Viewer) authoring and optimization
- Image targets on plaques; ARCore Geospatial API and VPS coverage checks; GPS fallback
- Offline caching, battery and thermal testing in sunlight
- App Store and Google Play release; App Clips / Instant Apps for low-friction entry
## Mind and voice
**AI / conversation engineer**
- Retrieval-augmented generation: chunking by dated entry, embeddings, hybrid keyword + vector search, citation IDs
- Policy layer: cutoff enforcement, refusal lists, trauma protocol, companionship ban; separate from the persona prompt
- LLM APIs (hosted) and a small on-prem model for offline or school-filtered venues
- Streaming speech: ASR (speech-to-text), streaming TTS, barge-in; end-to-end latency budgets
- Driving faces from audio: MetaHuman audio-driven animation or NVIDIA Audio2Face-3D (open source)
- Evaluation harnesses: golden question sets, regression on every corpus change, red-team suites
- Logging and dashboards (questions, corpus hits, refusals), privacy-safe analytics
**Conversation designer / narrative writer**
- Writing in period voice from primary sources; monologue structure
- Question-bank design (400–800 items) with citations and refusal flags
- Chautauqua format: in character, questions in character, out of character
- Accessibility writing: plain-language captions, children's cut
**Voice director / audio**
- Casting and dialect coaching; voice-bible sessions (calm, anger, humor, public address)
- Actor contracts for synthetic voice use; locking a TTS voice so it cannot drift
- Spatial audio, room tone, loudness standards for kiosks and headsets
## Research and rights
**Historian of record / corpus editor**
- Primary-source editing, transcription, citation standards
- Separating own words, contemporaries and period knowledge; writing D-notes (contested facts, silences)
- Blind review of generated answers
**Rights counsel**
- Collections and reproduction agreements; actor digital-replica clauses; right of publicity by state; trademarks
## Operations
**Kiosk / AV integrator**
- Windows kiosk hardening, auto-login, watchdogs, remote management
- Displays (portrait 55–86 inch, LED, holographic cabinets), microphone arrays with echo cancellation, depth or presence sensors
- Network: offline mode, captive Wi-Fi SSIDs, school allowlists
**QA**
- Device matrix testing (kiosk, Quest 3/3S, Vive, Vision Pro, iPhone, mid-range Android)
- Red-team prompt runs; caption accuracy; comfort testing
## Shared across the team
- Version control and asset naming conventions
- Reading a rights map before using any image or text
- The ground rules: [[Ground Rules — Historical Virtual Humans]]