# Capabilities and technologies: Ruslan Vasylev

Items marked "(proposed)" are designs, not delivered work.

## Capabilities

- Viewer UI: Inspect spatial information through selection, navigation and linked views.
- Clash detection: Identify mesh conflicts for general-contractor correction discussions.
- Review: Compare evidence, retain findings and communicate corrections.
- Remote presentation control: Sales-agent iPad controls drive property and unit views shown to buyers on a TV.
- 3D asset preparation: Turn source imaging, captured data and authored content into usable 3D assets.
- Responsive viewing: Keep loading, rendering and selected data aligned with interactive use.
- Property presentation: Guide buyers through property scenes and unit options.
- Unit selection and highlighting: Select property units and highlight them in the sales presentation.
- Camera presets and filtering: Apply camera presets and property filters through viewer state.
- BIM model loading: Load IFC data and assemble interactive browser models.
- Model editing: Prototype tools for vertex, edge and polygon interactions.
- Multi-selection: Select multiple model elements for interaction.
- Model transforms: Translate, rotate and scale model instances with viewer tools.
- Geometry measurement: Prototype library-backed distance, area, volume and angle measurements.
- Geospatial model inspection: View building and BIM data in GIS context for planning.
- Building-property editing: Edit building/model properties through browser interactions.
- Clash review: Review mesh conflicts and select findings for correction discussions.
- Local BCF issue creation: Convert selected clashes into locally retained BCF issues.
- Retain coordination findings: Save BCF topics, comments, viewpoints and element relations locally.
- BCF file exchange: Import and export coordination findings through BCF files.
- Organ selection and biomarkers: Select organs and show relevant biomarker information during doctor–member conversations.
- Volumetric imaging views: Inspect source slices and volume context alongside organ geometry.
- Prior/current scan comparison: Keep previous and current scan views aligned while explaining changes.
- MRI-to-mesh conversion: Convert imaging-team segmentation masks to browser-ready organ meshes.
- Mesh cleanup and levels of detail: Prepare validated geometry and multiple GLB detail levels.
- Geometry and placement validation: Check conversion outputs, transforms, scale and registration evidence.
- Reconstruction component selection: Use registration and reprojection-quality metrics to select usable components.
- Capture-to-3D reconstruction: Process source captures through frame extraction and reconstruction stages.
- Capture-to-asset delivery: Turn physical capture work into usable delivered 3D assets.
- Source traceability: Preserve source hashes, capture lineage and available placement evidence.
- Exact-version comparison: Review a specific reconstruction version and the evidence it used.
- Saved assessments: Retain assessments against their exact source, run, region and placement basis.
- Assessment corrections: Create explicit revisions linked to earlier assessments.
- Evidence sharing: Grant access to specific evidence and assessment briefs.
- Access revocation: Revoke shared access without rewriting the saved assessment.
- Processing progress: Show durable job state and meaningful progress separately from attempt ownership.
- Processing recovery: Recover stored completion reports and reconcile abandoned attempts.
- Resource-bounded processing: Apply input/runtime limits and fixed execution-profile checks before work runs.
- Processing diagnostics: Relate durations, iteration progress, errors and optional GPU samples to jobs.
- Versioned processing settings: Retain exact settings and check what execution actually applied.
- Result verification: Verify generated artifacts before committing completed job state.
- Cached model reuse: Reuse model bytes with version/checksum checks and invalidation.
- Selection consistency: Reject responses for an asset or reconstruction run that is no longer selected.
- CAD clearance checks (proposed): Proposed precise checks after conservative geometry candidate filtering.
- Revision-aware CAD findings (proposed): Proposed re-evaluation or explicit correspondence after model revisions.
- Customer-specific checks (proposed): Proposed configurable thresholds, materials, enabled checks and reference versions.
- ML-assisted review (proposed): Proposed use of versioned model inputs, predictions and human review.
- Model-change evaluation (proposed): Proposed holdouts, same-input shadow comparison, pinned rollout and rollback.

## Technologies used in the work

- Auth0
- AWS Batch
- BCF
- Blender
- Cache Storage
- Cesium
- CloudWatch
- COLMAP
- Docker
- EventBridge Scheduler
- Fastify
- FFmpeg
- GLB
- GLSL
- IFC
- IndexedDB
- JSON
- JSONB
- Lambda
- LichtFeld
- MQTT
- Node.js
- nvidia-smi
- Nx
- PostgreSQL
- Python
- React
- S3
- SHA-256
- Supabase
- ThatOpen
- three-mesh-bvh
- Three.js
- TypeBox
- TypeScript
- Web Workers
- web-ifc
- WebGL