One app for every leg of your trip — jeepney, bus, MRT, LRT, and walking — powered by MMDA data and on-device AI.
How do you usually get around?
KAirOS tailors your routes, parking, and coding alerts to how you travel. Pick one — you can change it any time in Profile.
Your route stays on your phone.
Routing stays local. With consent, only anonymous zones, departure window, and mode inform MMDA demand.
On-device AI
Anonymous demand
Offline ready
Location
For live navigation & arrival ETA
Notifications
MMDA incident alerts along your route
Microphone
Voice trip input (optional)
Home
Good morning, Gene 👋
Where do you need to be?
Trip Watch
Monitoring your planned trips
Forecasts are checked against MMDA road history, coding rules, and planner-published updates.
Know the trip before the trip
Plan a future departure. KAirOS watches the MMDA-aware forecast and lets planners see anonymous demand before it reaches the road.
Road closures · from MMDA
Planner-published closures that affect your area. Tap a trip to re-route around them.
Flood passability · MMDA scale
Confirmed reports and planner rules on the MMDA three-level scale. Routes obey each level per vehicle class.
PAGASA rainfall warning feed connects in the beta · illustrative here
Seen something on the road?
One tap sends flood depth, a crash, roadwork, illegal parking, a stalled vehicle or a dark signal to MMDA. Confirmed reports re-route everyone.
No open reports from this phone
MMDA travel window
Calculating best window…
…
Corridor speeds use MMDA Q1 2023 history. The hourly shape is illustrative because the source has no time-of-day column.
Station crowding
Shape from MMDA corridor speeds · per-station offset illustrative
MRT Cubao
Moderate · 7–9 AM peak
LRT-1 Baclaran
Low · best window 10 AM
BGC Bus Stop
Moderate · 5–7 PM peak
MMDA road network + historical travel time + UVVRP · demo data is labelled
Cubao
Ayala, Makati
Leaving now · 9:41 AM
Optimize for
MMDA-aware recommendation
Road network Travel-time history UVVRP Planner updates
Recommended now
38 min
Walk 5 min
MRT-3 (4 stops)
Walk 7 min
Leave by 7:42 AM Saves 17 min
Map and 2 alternatives
WalkJeepneyMRTBus
Safest
44 min
0 crash zones
Walk 6 min
LRT-1 → MRT-3 transfer
Walk 8 min
Recommended for night
Coding-safe
51 min
UVVRP-compliant
Walk 4 min
Jeepney · España–Lacson
Walk 9 min
Avoids plate-window
Same destination · two modes
Demo · illustrative fares & fuel
No active trip
Plan a trip from Home first, then KAirOS compares Drive vs Commute for you.
Re-route available.Crash on EDSA–Quezon Ave. Alt saves 6 min.
ETA
10:23 AM
14 min remaining
Smart mode-change announcement 0.31s
In 2 min, exit at MRT-3 Ayala on the north side. Platform is crowded — keep your belongings close. Walk 7 min along Ayala Ave to your destination.
Sa 2 minuto, bumaba sa MRT-3 Ayala, hilagang daan. Maraming tao sa platform — ingatan ang gamit. Maglakad ng 7 minuto sa Ayala Ave patungong destinasyon.
Generated on-device · context-aware · bilingual
Now
Walk 4 min
to MRT Cubao Station — use the north entrance
Next · bilingual
Board MRT-3 northbound
4 stops to Ayala · arrives 10:19 AM
Sakay ng MRT-3 patungong Ayala · 4 na hintuan
MMDA alerts
5 active incidents affecting your saved routes
On-device summaryNo data leaves phone
Heads up: a high-severity crash is blocking the southbound EDSA–Quezon Ave approach. Your Fastest route is affected — re-routing through Quezon Ave saves ~6 min. LRT-1 option adds 4 min but dodges all crash zones.
Paalala: malakas na crash sa EDSA–Quezon Ave. Naaapektuhan ang iyong Fastest route. Mag-re-route sa Quezon Ave, makatitipid ng ~6 min. Ang LRT-1 ay +4 min pero walang crash zone.
Filtered to your saved routes Processed on device · no data left phone
Your reports
Station crowding
Shape from MMDA corridor speeds · per-station offset illustrative
MRT Cubao · today
Best windows: 10 AM – 3 PM, after 8 PM
5a
6a
7a
8a
9a
10a
11a
12p
1p
2p
3p
4p
5p
6p
7p
8p
LowModerateHigh
Loading tip…
Reading the latest MMDA speed curve.
Current streak
7 days
142
Points
Badges
Off-peak Hero
Coding-Safe 7
Multi-modal
CO₂ Saver
30-day Streak
Unlock at 30-day streak
Safer Streets
Unlock at 5 crash zones dodged
This week
Off-peak hero
5 trips before 7 AM
0%
0 of 5 · done
Coding-compliant week
7 days without plate-window violation
0%
0 of 7 days
Multi-modal explorer
Use 3+ transport modes this week
0%
0 of 3 modes
Your impact
Last 30 days · personal TDM contribution
—
Trips planned
—
Crash zones avoided
—
Coding violations avoided
—
CO₂ saved
Your fuel savings · personal
Based on
—
Fuel not burned
—
Litres avoided
—
Km saved
Your saved minutes, translated to fuel and pesos — vehicle-aware (car or motorcycle). Demo · illustrative.
Trip time reduction
17%
You · 32 min avg
Metro Manila baseline · 38 min avg
On target for the 15–20% spec goal ✓
Behavioral change
Modal shift adoption
Chose public transit for a faster route
25%
Daily route queries
Avg of your logged trips · spec target 2–3
2.7
AI response time
Voice/text → route output on-device
1.4s
Departure shift
Avg departure vs the MMDA 8 AM peak · logged trips
—
Weekly trend
G
Gene Jr E. Gulanes
KAirOS · AI Developer & Web Engineer
How you get around
Plate & UVVRP
We automatically reroute around UVVRP-restricted roads on coding days.
Truck routes avoid narrow CBD cores and run at freight speed (set by MMDA in Routing policy).
Saved routes
Home — Cubao
38 min via MRT-3
Work — BGC
44 min via Bus
MMDA rules & enforcement
No Contact Apprehension (NCAP)
Camera corridors on your routes · official violation checker
Single Ticketing System
Uniform fines across 17 LGUs · what to do after a ticket
Preferences
Language
UVVRP coding alerts
Data & privacy
Notifications
No Contact Apprehension
Know where the cameras are before you drive. KAirOS keeps no violation records.
What NCAP is
MMDA cameras record traffic violations on selected corridors and the notice reaches the registered owner by post or through the official checker. No enforcer stops the vehicle.
Camera corridors on KAirOS routes
Illustrative corridor list from public MMDA advisories. In the beta MMDA enters and maintains the list from the console.
Most-cited violations
Check your plate
Look up any NCAP notice on MMDA's official checker, or in the eGovPH app. KAirOS opens the official page and stores nothing.
Content subject to MMDA approval before the beta · no personal violation data handled
Single Ticketing System
One ticket, one set of fines across Metro Manila's 17 LGUs. Works offline.
How it works
Since 2023 MMDA, the LTO and city enforcers issue one Single Ticket with uniform fines, so a violation in Pasig costs the same as in Makati, and a ticket can be settled in any participating LGU.
Uniform fines · common violations
Illustrative figures from public MMDA and MMC advisories. MMDA supplies and approves the final table before the beta.
After a ticket
Official channels
Pay or contest only through the issuing LGU, the MMDA, or the eGovPH app. KAirOS never collects payments.
Content subject to MMDA approval before the beta · stored on the phone for offline use
Plan · 1 of 15
Explore the mobile app
Overview
Demo clock · --:-- PHT
Aggregated · zone-level only
Low demandModerateHighMildModerateSeveremodelled severity · sample scenario Demand, safety, closure, and route lenses stay in place; severity colours show where a planned disruption hits hardest before it is published.
Network totals
—
Active commuters
Metro scope · estimated active
—
Trips today
modal shift to PT
—min
Avg trip time
avg · network trips
—
Modal shift to PT
PT share of scoped trips
Live activity · last 5 min
0 trips
Demo ticker · synthetic events streamed live during the demo so the dashboard doesn't look frozen. Real build would ingest MMDA operator logs.
Planned demand · scoped
—
Illustrative hourly shape from the labelled synthetic baseline · real MMDA corridor history lives in Observed.
Slowest corridors · real MMDA data
Q1 2023 · speed_kph
Corridor
Avg
Min
Max
Samples
Loading slowest corridors from MMDA Travel Time Data…
Road network coverage · by city
18 cities · 10,260 segments
City
Segments
kph
%
Loading coverage from MMDA Travel Time Data…
AI vehicle detection · on-device
Review this operational signal alongside planned Trip Watch demand and historical TMC/AADT records. A real YOLOX-Tiny model detects & boxes vehicles in-browser via ONNX Runtime Web on any traffic clip you load. No footage ships with this demo and no live MMDA feed is used. Plate reads and vehicle counts are simulated. A clip is decoded on this device and never uploaded.
AI: loading…
Footage
none bundled
Drop a .webm or .mp4 traffic clip on the stage, or choose one. It is decoded and counted on this device and never uploaded. Running the model on the CCTV footage MMDA supplies is an optional upgrade, outside the beta's core scope; this demo carries none.
No clip loadedAny traffic clip you load · counted on-device by YOLOX-Tiny
awaiting clip
REC
NO CLIP
2026-07-16 17:42:00
Load a traffic clip
Drop a .webm or .mp4 here, or click to choose one. The model boxes vehicles on this device; nothing is uploaded.
AI Model: loading…Inference —— FPSSession 00:00
Live vehicle count · no clip
counting
0vehicles · 0/min
Vehicle counts are simulated for demo stability. The detection boxes on a loaded clip are the live on-device YOLOX-Tiny output — or a simulated fallback if the model is unavailable.
Plate reads · simulated
Demo · not real ALPR
reading0 coding violations
Load a clip to start reading plates
Simulated plate reads for this demo — generated locally, never real plates. Each read is checked against today's UVVRP coding schedule below.
Today
UVVRP coding
—
Plate endings restricted
AI-assisted detection pipeline (roadmap)
Roadmap · aspirational
License plate recognition — UVVRP violation flagging at camera positions (this demo shows a simulated version)
Speed analysis — cross-reference with Travel Time Data for ground-truth calibration
Incident detection — auto-detect crashes, stalls, and flooding; trigger commuter alerts + tow dispatch
Choose a sample scenario — the affected road appears on the map and the impact widget docks beside it. Nothing reaches commuters until you publish.
Temporary response
Should KAirOS publish a temporary route update for EDSA roadwork?
Model the planned disruption, review the affected demand, then decide whether to update matching Mobile trips.
Recommended response after impact review
Publish the closure and reroute matching Mobile tripsPlanner approval is required before anything changes in Mobile.Ready to compare
Define Compare Publish Monitor
Adjust disruption details
Reference network
No disruption
Avg trip time
—
Affected commuters
—
PT modal share
—
Predicted complaints
—
Planned disruption
Modelled effect
Avg trip time
—
Affected commuters
—
PT modal share
—
Predicted complaints
—
Illustrative scenario over the labelled synthetic trip baseline. Publishing writes a planner update to this browser only; it is not a live MMDA incident feed.
A published closure is a Close rule: while it is in force, every route search in this browser removes the road for all vehicles. Rules steer KAirOS users only.
Monitoring · Mobile updates
0 active
Awaiting verification
0 pending
Each report needs two road-user confirmations or one console check. Confirming publishes the matching MMDA rule for the report's expiry window; dismissing drops it.
Triage: loading…
A decision model reads each report's note and answers fixed questions: is this about a road, and does it fit the hazard the user tapped. When it reads a note as junk, the note's text is hidden from commuters' Alerts until a planner checks it; the report itself still shows. It is a classifier, not a text generator. It orders this queue and a planner confirms or dismisses every report. In tests it read English notes well and Tagalog notes poorly, so an unsure answer means check, not doubt. Answers for the sample notes were computed offline; any other note gets a keyword check unless a local decision service is connected.
Confirmed · routing applies
0 active
Reports come from KAirOS users in this browser; sample reports are labelled synthetic. In the beta, reports also arrive from other phones through the API. MMDA advisories are keyed in by console staff at launch; a feed of agency posts on X joins the same queue later in the experiment.
Social-experiment KPIs · Jan–May 2027 design
This browser · synthetic cohort
Each behaviour KPI compares nudged trips with the stage-1 baseline and a 10% holdout that sees forecasts but no incentives. Columns: what this device has logged, and the labelled synthetic cohort used to design the dashboard.
Departures by hour · synthetic cohort
peaks shaded
Time-shift counts nudged departures that land outside the shaded peaks. The cohort is the labelled 5,000-trip synthetic baseline, not experiment data.
Share the data
The monthly KPI report to JET starts from this CSV. The GeoJSON is the same payload the console publishes through the REST interface in the beta; here it is built from this browser's rules.
Standing rule
Standing rule: how KAirOS routes cars and trucks
This rule applies to future matching Mobile route searches until a planner changes it.
Active rule
Loading planner-set policy…Demo policy · stored in this browserActive
Models slower truck movement from weight and congestion
Comma-separated · matched against road names
Car routing
1.00
Cars stay on the full drive graph with UVVRP compliance
MMDA routing rules
0 in force
Four effects, applied in this order: Close, Restrict, Divert, Weight. Close and Restrict are absolute. Divert and Weight are advice: Mobile shows MMDA's route beside the fastest legal one, within the detour limit.
Suspend all advice rules
Divert and Weight stop steering routes. Close and Restrict stay in force.
Draft a rule
60%
2.0×
Rule log · publishes and withdrawals
Planner-published in this browser, not a live MMDA feed. Rules steer KAirOS users only. The preview routes the labelled synthetic seed trips, not real demand.
Route effect · Cubao → BGC
Current active rule
Car is the reference route. Truck shows the standing rule applied by the browser router. Both obey the MMDA routing rules in force now. Planner-set demo values.
Data & provenance · 8 loaded datasets + 2 live sources
AvailableRoad-user reports0 reports · this browser · verified in Reports
PendingPAGASA warningsPublic rainfall + thunderstorm feed · connects in the beta
Companion pipeline tool that converts raw MMDA datasets into the
lightweight JSON the app loads. Static view — run
python3 scripts/build_data.py --status for the live report.
The production system the PP2 beta runs on: two clients, one API, one routing engine, one database, all built once from Terraform and scaled by rules, so after handover only the code changes.
Planned, not deployed. This demo is fully client-side (static JSON, browser routing, LocalStorage). The page below is the beta design from Technical Proposal 3.5 and docs/INFRASTRUCTURE.md, with the sizing model from that plan. Every figure is an October 2026 estimate for the Inception Report, not a measurement.
Route requestCached read (the burst path)Telemetry, queuedRule → push fan-outInternal callExternal system
Set up once, then only code changes
Terraform builds the whole account during Phases 2 and 3 and is then left alone. Code reaches production one way: merge to main, GitHub Actions builds the images, ECS rolls them out with a circuit breaker. Road-network refreshes are a commit to the tile build, never a server change.
MMDA touches: the repository, the CloudWatch dashboard, the bill, snapshot restore, secrets rotation, two Terraform variables at most.
MMDA never touches: task definitions, scaling policies, security groups, the EC2 instance, SSH (there is none), the console by hand.
The one scaling case: the push burst
With 5,000 users, steady traffic is under one route a second. When MMDA publishes a flood rule, thousands of phones open within two minutes: 50 to 100 times steady. Fargate takes 2 to 3 minutes to add a task, so the burst is absorbed before the API, and the API is raised before the push goes out.
Forecast and advisory reads are cacheable GETs with ETags, and every phone asks for the same URLs (5-minute slots, one advisory version), so CloudFront serves the repeats.
The publish is a console action, so the worker raises the api floor to 4 tasks and staggers the zone topics before any phone is pushed.
Phones wait a random 0 to 90 s after a push before refreshing, and re-route only if a saved trip touches the rule's segments.
WAF limits by device, not by address: hundreds of phones share one carrier-NAT address.
Scaling rules per tier
api 2 to 8 tasks. Target tracking on ALB requests per target, set from the January load test, and CPU 60%; a step adds 2 in a burst. Out in 60 s, in after 10 min.
worker 1 to 4 tasks. Target tracking on backlog per task (telemetry messages ÷ tasks, target 300); one poll loop per queue.
routing 1 instance, ceiling 2 only so a deploy starts the new task before the old stops. Grows by changing one instance-type variable.
decide (report triage) a Lambda function on the triage queue; scales to zero by itself.
database no compute autoscaling; storage grows to 100 GB by itself. Multi-AZ and a read replica are flags, off.
The three flows
Route. Phone → ALB → api: rules in force at departure become Valhalla exclusions per vehicle class; Divert and Weight rules are applied in the api over Valhalla's alternatives, as the prototype's router does. Target 1.5 s at p95.
Telemetry. A position every 5 s, uploaded each minute, trimmed 300 m at each end, answered 202 and queued. The worker map-matches and writes segment speeds. Raw positions are never stored.
Report → rule → push. Two user confirmations or the console publish the rule, credit 5 points and fan out a push scoped to the corridor.
Security and RA 10173
Phones carry a random device ID and a bearer token; no accounts, phone numbers or emails.
Console users have passwords (argon2id) and TOTP; every write lands in the audit log.
One IAM task role per service; deploys through GitHub OIDC, so no long-lived AWS keys exist.
TLS 1.2+ at the edge; RDS, EBS, S3, SQS and logs encrypted at rest; secrets only in Secrets Manager.
Zone-level intents, 80-character report notes filtered for plates and numbers, plate suffix kept on the device.
Budget and milestones
Production is quoted at USD 255 a month: database 85, Fargate 40, routing instance 65, load balancer 35, storage and transfer 15, queues and monitoring 15. Staging adds 40. Budget alarms at 200 and 300. The slider follows the plan's growth path: about 380 at 20,000 users, 900 at 100,000, 1,750 at 500,000 and 2,500 at 1,000,000; the plan's three further changes (segment matching on the phone, one upload at trip end, reservations bought by MMDA) would bring 1,000,000 to about 1,500.
Mid Nov 2026 staging from Terraform; CI plans every PR.
Dec 2026 tiles from OSM + Annex A; first load test.
Early Jan 2027 production, restore test, 5,000-device burst test under 1.5 s p95.
15 Jan 2027 beta on the stores. 1 Jun 2027 cost report from real data. Aug 2027 account handed to MMDA.
Trade-offs chosen on purpose: ECS tasks sit in public subnets behind security groups instead of paying about USD 35 a month for a NAT gateway (a private_tasks flag adds one); the database is single-AZ with point-in-time recovery (a 5-minute recovery point, where the proposal promises a day) and the routing instance is one, replaced in about 5 minutes, because during either outage the app plans saved trips on the device; Valhalla has hard exclusions but no soft per-area cost, so advice rules are scored in the api. Full detail: docs/INFRASTRUCTURE.md.