RelayOne AIRelayOne.AI Launch Live Demo
Prototype · Multilingual dispatch across India

When every second counts,
no caller is left on hold.

RelayOne answers emergency calls in the caller's language, structures the incident, locates it on a map, and hands the dispatcher one screen to act on. All in under two seconds.

Launch Live Demo Explore Dashboard
relayone / operations · hyderabad-fire-2041
Live · 12 active calls
Hyderabad
Chennai
Kolkata
Lucknow
Priority
Critical High Medium Low
Incident
Factory fire, workers trapped
Location
Balanagar Industrial Estate, Hyderabad
Caller
Telugu, translated to English
AI confidence
94%

Dispatch Timeline

INC-2041
  1. 00:00
    Call answered in Telugu
    Voice captured, streaming ASR started
  2. 00:00
    Language detected
    Telugu (TE), confidence 0.98
  3. 00:01
    Incident classified
    Structure fire, victims trapped upstairs
  4. 00:01
    Location resolved
    Balanagar Industrial Estate, Hyderabad
  5. 00:02
    Brief handed to dispatcher
    One-screen brief, action recommended
  6. 00:04
    Units acknowledged
    2 fire tenders, 1 ALS, 1 police unit

Resource Status

4 assigned
  • Fire Tender FT-04ETA 3m
  • Fire Tender FT-07ETA 4m
  • ALS Ambulance A-12ETA 5m
  • Police Unit P-21On scene
1
On scene
3
En route
8
Available
The Response Pipeline

One continuous flow, from first ring to first responder.

Nine stages, one system. Each stage feeds the next in real time.

  1. 01
    Emergency Call
    Caller dials, no app required.
  2. 02
    AI Answers
    First response in under a second.
  3. 03
    Speech Recognition
    Streaming ASR with word timing.
  4. 04
    Language Detection
    13+ Indian languages, auto-detected.
  5. 05
    Understanding
    Claude reasons over the caller's story.
  6. 06
    Translation
    Dispatcher reads English in real time.
  7. 07
    Location
    Address, hazards, nearest units.
  8. 08
    Dispatcher
    One brief, one screen, one decision.
  9. 09
    Units Dispatched
    Fire, police, medical, en route.
End-to-end latency target< 2 seconds
AI stage Human touchpoint
System Architecture

A three-stage engine that turns a panicked voice into a dispatch order.

Input, processing, output. Every module has a job and a defined signal it hands to the next stage.

Stage 01
Input
Signals in
  • Citizen
    The person on the line, often under acute stress.
    Outputs:Raw voice stream
  • Voice
    Continuous audio capture with word-level timing.
    Outputs:Timed audio frames
  • Language
    Detects the language before transcription completes.
    Outputs:Language code + confidence
  • Location
    Cell signal and caller phrasing seed geolocation.
    Outputs:Approximate coordinates
Stage 02
AI Processing
Reasoning loop
  • Speech Recognition
    Streaming ASR transcribes the caller in real time.
    Outputs:Native-language transcript
  • Translation
    Two-way translation so the dispatcher reads English.
    Outputs:English transcript
  • LLM Reasoning
    Claude reads the story and asks the right follow-ups.
    Outputs:Structured incident record
  • Classification
    Maps the incident to a response taxonomy.
    Outputs:Type, sub-type, tags
  • Hazard Detection
    Cross-checks the scene against chemical, fire, flood layers.
    Outputs:Hazard list + exclusion zones
  • Priority Scoring
    Severity, vulnerability, and elapsed time set the score.
    Outputs:Priority 1 to 4
Stage 03
Output
Actions out
  • Dispatcher Dashboard
    One screen with brief, map, recommended units.
    Outputs:Actionable brief
  • Police
    Nearest patrol units notified with route and role.
    Outputs:Patrol assignment
  • Fire
    Tenders and hazmat crews staged with scene notes.
    Outputs:Fire response order
  • Medical
    Ambulances routed with victim count and injuries.
    Outputs:Medical response order
Claude
Reasoning
Deepgram
Streaming ASR
FastAPI
Realtime services
Redis
Live state
Mapbox
Geospatial
WebSockets
Sub-second updates
Next.js
Dispatch console
LoopDownstream signals feed back into the LLM's next turn.
HumanThe dispatcher confirms every action. AI never dispatches alone.
Why It Matters

Built for the scale of a country in motion.

22
Official languages spoken across India
40M+
Emergency calls placed in India each year
8 min
Average hold time during a monsoon surge
< 2s
AI response time
13+
Languages supported
90%
Triage accuracy
100%
Address extraction
24/7
Always available

An AI & Robotics Hackathon prototype of India's next-generation emergency command platform.

Launch Live Demo