{ }

See Everything.
Understand Everything.

Can machines truly understand what happens in video? We are building the answer — one domain, one deployment, one hard problem at a time.

Capabilities
How it works

Four stages. One pipeline.
Each one taught us something we did not expect.

01

Ingest

Decode any video source. Extract frames at adaptive sampling rates that respond to scene complexity. Chunk into overlapping segments — so events at chunk boundaries are never missed.

  • MP4 / MKV / RTSP / HLS
  • Adaptive 1–10 FPS sampling
  • Overlapping chunks prevent boundary loss
02

Analyze

Two AI systems cross-validate every chunk: vision language models understand context and meaning; YOLO catches precise objects and locations. Results are fused — fewer false positives, fewer missed detections. Every output is schema-validated before it reaches your systems.

  • VLM + YOLO cross-validation
  • Zod schema enforcement
  • Automatic multi-provider failover
03

Understand

Aggregate chunk-level data into video-wide intelligence. Events spanning multiple chunks are correlated and deduplicated. Entities are tracked across cameras using privacy-preserving text descriptions — no facial recognition required.

  • Cross-chunk event correlation
  • Privacy-preserving entity tracking
  • Temporal deduplication engine
04

Act

VideoRAG powers natural language search across your entire archive — with timestamp-accurate cited answers. Define alert rules in plain English. Get structured reports, API access, and real-time notifications.

  • VideoRAG semantic search
  • Natural language alert rules
  • REST + WebSocket API
Use cases
Domains

One question, many proving grounds.

Each domain tests a different facet of the same core question: can machines understand what happens in video well enough to act on it? Every deployment sharpens the answer.

[SEC]

Security

Perimeter awareness. After-hours activity. Abandoned object alerts. Faster incident review.

Learn more →
[WHS]

Warehouse

Dock visibility. Staging-zone congestion. Forklift safety context. Throughput exceptions.

Learn more →
[DEF]

Defense

Drone surveillance analysis. Convoy tracking. Situational awareness from aerial feeds. Force protection.

Learn more →
[MFG]

Manufacturing

Quality inspection via visual AI. Assembly line monitoring. Safety compliance. Equipment anomaly detection.

Learn more →
[RET]

Retail

Shelf analytics. Customer flow heatmaps. Loss prevention. Queue management and wait-time estimation.

Learn more →
[TRN]

Transportation

Traffic flow analysis. Incident awareness. Roadway disruption monitoring. Pedestrian safety signals.

Learn more →
[HLC]

Healthcare

Patient safety awareness. Care-environment monitoring. Waiting room visibility. Privacy-aware review.

Learn more →
Developer platform
For builders

Shared tools.
Serious problems.

REST APIs. `@vii/app-sdk`. `@vii/ui-kit`. Everything we build is accessible programmatically and composes from the same platform UI system.

  • Full REST API with OpenAPI spec
  • App SDK for platform-native apps
  • Shared UI kit and shell tokens
  • Webhook notifications for downstream workflows
  • Self-hosted deployment paths for controlled environments
analyze.tsui-kit.ts
// Submit a video, then poll for results
import { useVideoAPI } from '@vii/app-sdk'
import { Button } from '@vii/ui-kit'

function AnalyzePage() {
  const api = useVideoAPI()
  const run = async () => {
    const job = await api.submit(
      'https://storage.example.com/security-cam-north.mp4',
      { sampleRate: 2, appId: "security" }
    )
    const status = await api.getStatus(job.jobId)
    const events = await api.getEvents(job.jobId)
    const summary = await api.getSummary(job.jobId)
  }
  return <Button onClick={() => void run()}>Analyze Video</Button>
}

// job → { jobId, status, statusUrl, detailUrl }
// status → { status, progress }
// events / summary fetched after processing completes
Shared SDK hooks and manifest contracts, documented end to end✓ type-safe

The question is open.
Help us answer it.

We are looking for partners with hard video problems and the patience to solve them properly. Deploy in your environment. Keep your data private.