Ideas & Proposals for
Diriyah Smart City
A curated set of ideas, designs, and working prototypes — proposed as advisory input to Diriyah's transformation under Saudi Vision 2030. From board-level strategy to real-time operations, from the visitor's phone to the backend that makes it all work.
What This Is
Not a product pitch. A set of concrete proposals — each one a video, each one addressing a real question Diriyah leadership, operations, or visitors will face. The proposals span three dimensions:
Strategy & Outcomes
How to measure what matters — 6 business KPIs connecting visitor experience, city image, safety, revenue, VIP retention, and global ranking to concrete drivers. Storyline A.
Experience & Engagement
How a visitor's day is transformed — from arrival through heritage touring, dining, weather response, VIP evening, to departure. Plus the mobile app putting AI in every visitor's pocket. Storylines B & D.
Intelligence & Operations
What powers the above — a 14-zone command center, crowd prediction and stress testing, and the backend architecture from sensor data to sub-500ms predictions. Storylines C, E & F.
HDiriyah Super AppOpen Super App →
- Civic Pulse · today in Diriyah at a glance
- Bills & Services · one-tap pay
- Citizen Reporter · snap → submit
- Local Loyalty · points · redeem
- Ambassador Program · paid local guide
- Concierge · human · 24/7
- Gate access · timed entry · QR
- Heritage tour · guided routes
- Family & etiquette guidance
- My Diriyah Story · auto-cut · share-ready
- Trilingual UX · EN / AR / 中文
- Visa & arrival assistant
- Cultural etiquette primer
- Currency & tipping helper
- "Could I live here?" lifestyle preview
- Dedicated RM · 24/7
- Helicopter ops · chauffeur dispatch
- Private Salwa Palace tour
- Ghost mode · discreet routing
- Family Office & HSM-grade messaging
- Catalog & bookings
- Live demand signals from the city layer
- Payouts & settlement
- Ratings & reviews
- Cross-merchant promotions
- Investor brief · curated by RM
- Private property tour booking
- Construction & project transparency
- Pipeline tracker · attribution (90d)
- Closing room · documents & e-sign
GOrchestration Console
G-0: Gate B Hotspot — Full Decision Cycle
~2:37A complete orchestration cycle in real time: crowd surge at Gate B triggers detection, recommendation, operator approval, coordinated four-channel fan-out, and automatic rollback.
- Console Layout: live scenario state (left), Diriyah event zone map with 3 gates and density percentages (center), recommendation queue (right)
- Trigger: Gate B utilisation climbs to 75%, red pulsing halo — hotspot confirmed, rules engine evaluates
- Recommendation: EVT-HOT-001 — divert 30% from Gate B to Gate C, 90 vehicles, 3 minutes saved. VMS, ambassador, and app push
- Fan-out: VMS billboard flashes "Gate C, 3 min faster" → OPS-2 dispatched to Gate B → golden divert arrow appears — one approval, four channels
- Cooling & Rollback: Gate B drops from high-80s to ~60%, VMS reverts, staff returns, arrow fades — baseline automatically restored
- Auditability: full timestamped trail of every tick, rule, approval, and rollback; KPIs: latency, quality, channel parity
G-1: Parking Cascade — System-Level Rebalance
~3:07Unlike a single-gate hotspot, this is system-wide saturation — three gates filling in waves. The operator faces two simultaneous recommendations and chooses the strategic response.
- Cascade Trigger: Gate A rises first (wave 1), then Gate B follows (wave 2) — both exceed 80%, the cascade rule fires
- Two Concurrent Cards: EVT-HOT (local fix: 90 vehicles, 1 VMS, 1 staff) vs PARK-CASCADE (system rebalance: 240 vehicles, 2 VMS, 2 staff, overflow lot)
- Operator Choice: the strategic cascade rebalance is approved — not the local patch
- Four-Channel Fan-out: VMS North "Park at P-Overflow, 5 min" → VMS East "P2/P3 recommended" → OPS-1 to P1 → OPS-3 to P-Overflow
- System-Level Cooling: all three gates cool together — total pressure reduced, not just shifted
- Synchronized Rollback: both VMS signs revert, OPS-1 returns to Bujairi, OPS-3 moves to Ceremony Plaza
G-2: Zone Density — Soft Intervention
~2:35No crowd surge, no gate crisis — just a visitor zone getting too full. The system responds with a gentle app push, not a forced divert. Watch the colour change from gold to blue.
- Zone Density Trigger: Bujairi Terrace crosses 85% — the density threshold fires, but priority is Medium, not High
- Soft Channel: App push only — no VMS, no staff dispatch. 130 listeners, divert 10%. A suggestion, not an order
- Blue vs Gold: the divert arrow is blue (suggestion) instead of gold (directive) — the orchestration layer matches response to severity
- Gradual Migration: Bujairi slowly drops, Overlook Deck slowly rises — ~13 visitors respond (~10% conversion)
- Auto-Rollback: Bujairi drops below 75%, blue ripple and arrow fade — no manual intervention needed
- Design Principle: three scenarios, three intervention types — the system doesn't treat every problem the same way
FCrowd IntelligenceLive Demo →
F-0: Peak Alerts & Crowd Prediction
~3:12Follow the data trail from an Exhibition Hall red alert through prediction models to the exact moment a diversion decision must be made — and what happens if it isn't.
- Live Dashboard: 58,500 on-site (70% of 83K capacity), 2 zones in alert, peak predicted at 8 PM — a 4-hour action window
- Zone Analysis: Exhibition Hall at 93% (red), approaching 4 people/m² crush threshold within 30 minutes
- Gate Imbalance: South Gate at 96% with 22-min wait, while North/East Gates run at 53-64% — data basis for diversion
- Prediction Model: Actual numbers exceeding forecast by 15% — tonight's real peak could hit 13,000, a thousand over safe capacity
- Heatmap Intelligence: Redirecting 15% of Exhibition Hall inflow to adjacent zones keeps all areas within safe limits
F-1: Stress Testing & Scenario Simulation
~3:18Two stress test scenarios back to back — 60K vs 150K visitors — revealing the precise safety threshold where crowd management shifts from manageable to critical.
- Diversion in Action: South Gate reroute saves 10 minutes per visitor (22 min → 12 min), redirecting 2,000 visitors/hour
- 60K Scenario: Safety grade B, peak 83% capacity, 9-min avg wait, 54-min evacuation — globally safe but 2 zones locally stressed
- 150K Scenario: Safety grade D, 176% capacity, 38-min avg wait, 143-min evacuation — every zone a bottleneck, scenario not viable
- Side-by-Side Comparison: Risk score jumps from 40 to 90/100, evacuation time nearly triples — quantified gap between safe and dangerous
For every 10,000 additional attendees, the system calculates remaining safety margin, first failure point, and required countermeasures.
EBackend Architecture
E-0: Crowd Intelligence Platform — Backend Architecture
~18:50A 46-slide deep dive into the backend systems behind the crowd prediction, dynamic diversion, and stress testing demos (WP-3.3.3 & WP-3.3.4). The full chain from data collection to real-time inference to simulation and operations.
- Data Ingestion: 7 source categories (IoT sensors, computer vision, WiFi/BLE probes, ticketing, mobile GPS, social, weather) unified through a Kafka message bus with stream (Flink) + batch (Spark/Delta Lake) processing
- Feature Store: 120+ features across temporal, spatial, external, and behavioral categories — offline Spark pipelines for training, online Flink CEP for real-time inference (<3s end-to-end)
- Three-Tier Model Strategy: Temporal Fusion Transformer for 0-4h predictions (MAE <500), XGBoost for 4-24h trends, Prophet for day-to-week planning, plus Graph Neural Networks for inter-zone flow distribution
- Serving Infrastructure: Kong/Nginx API gateway → TorchServe/Triton model cluster → Redis cache → WebSocket push. P99 latency <500ms, throughput >1000 req/s, 99.9% availability
- Alert & Diversion Engines: 4 escalating alert types (threshold, trend, anomaly, composite); constraint-based optimizer generates route/gate/queue recommendations in real time
- Simulation Engine: Discrete Event Simulation + Agent-Based Model hybrid, 10,000+ scenario combinations across 6 parameter dimensions, triple-method bottleneck detection, evacuation target <60 min (Saudi Civil Defense compliant)
- Hybrid Cloud-Edge Deployment: Training in the cloud, inference at the edge on NVIDIA Jetson nodes. 30-minute offline autonomy with automatic failover and data backfill
- MLOps & Operations: DVC + MLflow for versioning, champion-challenger retraining (scheduled + drift-triggered + new-event-triggered), Kubernetes orchestration, Prometheus/ELK/Jaeger observability stack
- Security & Compliance: SHA-256 MAC hashing at collection, AES-256 at rest, TLS 1.3 in transit, OAuth 2.0 APIs, PDPL (Saudi Data Protection Law) and ISO 27001 compliant
- Delivery Roadmap: 3 phases over 12 months — PoC (months 1-3), Pilot (4-6), Full Production (7-12)
DVisitor AppLive Demo →
D-0: Diriyah Visitor App
~2:35A mobile-first H5 app putting five AI-powered tools into every visitor's pocket — Smart Parking, Timed Entry, Dining, Heat Safety, and AI Navigation.
- Smart Parking: Real-time availability for Samhan Parking (1,700 spots) and Diriyah Square (10,500 spots), floor-by-floor color-coded capacity
- Timed Entry: At-Turaif UNESCO site — 200 visitors per 15-min slot, instant QR code ticket generation
- Dining: 28 Bujairi Terrace restaurants with live wait times, AI cross-restaurant suggestions when venues are full
- Heat Safety: Real-time temperature, 3-level safety gauge, 6 cooling stations with walking distances, Emergency SOS
- AI Navigation: 14-zone SVG map, walking time calculation, trilingual AI chat (English, Chinese, Arabic)
Includes full Arabic RTL layout demo and Chinese language support. The same AI intelligence powering the Command Center, delivered to 50 million annual visitors.
CLive OperationsLive Demo →
C-0: Live Operations Command Center
~3:08The complete operations walkthrough — real-time monitoring, AI prediction, crowd simulation, facility management, and trilingual support.
- Command Center Overview — 14-zone map, live visitor count (66K+), zone ranking
- Flow Prediction — actual vs AI prediction, 94.2% model accuracy
- Crowd Surge Simulation — Bab Samhan gate surge, 5 AI response steps
- Facility Failure — HVAC failure affecting 8,200 visitors, Maximo auto-dispatch
- IBM Maximo Integration — infrastructure health dashboard, auto work orders
- Arabic Language Demo — full RTL layout switch
- Return to Strategic View — closing the loop
BVisitor Experience JourneyLive Demo →
B-0: Journey Overview & Arrival
~2:27Full day-at-Diriyah timeline — six touchpoints from 9 AM arrival to 8 PM departure. Deep dive into Arrival at Bab Samhan Gate.
- WITHOUT: Single gate, 10,500 parking spots bottlenecked, 90% density, 25-min queue
- WITH: Three gates activated, distributed parking, 20% density, 4-min entry
- AI Flow: 2-hour arrival prediction → smart parking routing → multi-gate dynamic assignment
Role perspectives: Operations sees throughput; VIP Manager sees valet pre-staging; Security sees density — same moment, three dashboards.
B-1: Heritage Tour & Dining
~2:16Two touchpoints: Heritage Tour at At-Turaif (10:30 AM) and Dining at Bujairi Terrace (1:00 PM).
Heritage Tour: 68% density → 42% AI-balanced flow, timed entry protecting UNESCO site. Crowd sensors at fragile zones → 200/15min slot limit.
Dining: 100% full at 1 PM with 45-min waits → demand spread across 3-hour window, 8-min wait. 2-hour forecast → unified booking across 28 venues.
B-2: Weather, VIP Evening & Departure
~2:27Three touchpoints completing the visitor journey: Weather Emergency, VIP Evening at Opera House, Smart Departure.
- Weather (3 PM): Zero incidents — 200 IoT sensors trigger at 39°C with 90-min protective window
- VIP Evening (6 PM): AI guest profiling → curated Opera House program, 94% satisfaction
- Departure (8 PM): 35-min exit → 8-min via wave scheduling and shuttle surge
AStrategic GoalsLive Demo →
A-0: Strategic Overview & Visitor Experience
~2:24The opening video. Introduces the Command Center's strategic dashboard with all six KPI cards, then dives into Visitor Experience — the most comprehensive KPI.
- Bab Samhan Gate Wait Time: 25 min → 8 min via AI multi-gate distribution
- At-Turaif Timed Entry: 200 visitors per 15-min slot protecting UNESCO heritage
- Bujairi Terrace Dining: 45 min → 12 min wait via time-spread recommendations
- Wadi Hanifah Heat Comfort: 3 incidents/month → 0 via proactive IoT response
Concludes with the What-If Simulator showing system performance at scale (up to 150K daily visitors).
A-1: City Image Score
~1:54How Diriyah climbed from B+ to A in international benchmarks. Radar chart comparing against Dubai Expo, Singapore, and NEOM.
- UNESCO Heritage Protection: AI crowd management earned audit upgrade B → A
- Bujairi Terrace Visibility: Media features 3 → 28 in 6 months
- Global Media Sentiment: 62% → 78% positive across 40+ languages
A-2: Safety Index
~1:4796.2% safety with zero major incidents across 180 days. Monthly incident trend and breakdown.
- Salwa Palace Crowd AI: Camera-based counting across 3 historic quarters, stampede risk monitoring at 4 narrow Najdi corridors
- Wadi Hanifah Heat AI: 200+ IoT sensors across 120km valley, AI triggers at 39°C — response shifted from +45min reactive to -90min proactive
A-3: Revenue Impact
~1:50+8.3% YoY revenue growth, $7.2M additional per-capita visitor spending. Revenue breakdown by category.
- Bujairi Terrace Turnover: 1.2x → 1.5x table turnover via AI time-spread and sunset-premium reservations
- Diriyah Square Retail: AI-guided discovery across 400+ units, per-visitor spend up 15%
Includes the revenue What-If Simulator showing the gap between AI-assisted and unassisted operations.
A-4: VIP Retention
~1:5164% return rate, up from 54%. VIP conversion funnel and quarterly retention trend.
- Bab Samhan Hotel Personalization: AI guest profiling, 23 unique Najdi suites matched to guest personality, return booking 54% → 64%
- Private Salwa Palace Tours: 30-min private window, AR heritage overlay, private sunset from rooftop (94% satisfaction, 82% rebooking)
A-5: Global Rank
~1:22From #8 to #5 in global smart city benchmark. 8-dimension radar chart — Diriyah leads Heritage (95), Culture (92), and F&B (88).
Dimension ranking table with head-to-head comparisons against Dubai, Singapore, and NEOM. Ranking progress chart shows acceleration in the most recent quarter.