🏗️ CIVORA
Every site decision, safer, leaner, and lower-carbon. 🌍
The AI command center that turns fragmented construction data into one accountable decision loop.
🌍 Inspiration
"10:22 AM. Six workers. 18 minutes from a real problem."
Construction sites already generate enormous amounts of data — environmental sensors, inspections, workers, tools, schedules, cameras, material records. Yet the people responsible for urgent decisions still get that information through disconnected dashboards, spreadsheets, and radio calls.
This isn't a small problem:
| 🌐 Global Impact | 📊 Stat |
|---|---|
| Built environment's share of global energy use | 32% |
| Built environment's share of global CO₂ emissions | 34% |
| Industry risk ranking | One of the world's most hazardous |
Safety teams spot hazards without full context. Project leads see delays after productivity is already lost. Sustainability reporting happens long after the decisions that mattered were made.
We kept coming back to one moment: six workers approaching an unsafe heat-exposure threshold in a corridor. The temperature alone isn't unusual — but combined with poor ventilation, task intensity, accumulated exposure, and fatigue, they're 18 minutes from a real problem. Most sites only discover this in hindsight.
So we built something that catches it in the moment, explains why it matters, and helps someone act on it — accountably.
🔁 That became Civora's core loop:
🔍 SENSE → 🧠 UNDERSTAND → 💡 RECOMMEND → ✅ APPROVE → ⚡ ACT → 📋 VERIFY
🚀 What It Does
Civora is an AI-powered construction operations command center for site managers, safety leads, sustainability teams, workforce coordinators, and tool managers.
It doesn't just display information. It prioritizes risks, explains contributing factors, recommends practical controls, routes accountable actions, and records what changed.
🧩 Seven Connected Workspaces — One Platform
| Workspace | Icon | What It Delivers |
|---|---|---|
| Command Center | 🧭 | Live site safety score, KPIs, AI-prioritized intervention cards, persistent tasks, activity stream, modeled monthly value |
| Live Digital Twin | 🗺️ | Spatial zones with risk/workforce/environmental/asset layers, live telemetry drift, 2-hour risk forecasting, privacy-preserving edge-vision demo |
| Safety Intelligence | 🛡️ | Prioritized risk register, explainable factor breakdowns, confidence scores, human-in-the-loop approvals, field observations, PPE scans, drills, multilingual briefings |
| Carbon & Circularity | 🌱 | Carbon budgets, baseline vs. AI-optimized scenarios, waste diversion, digital material passports, circular material matching |
| Workforce Intelligence | 👷 | Crew readiness, competency coverage, fatigue-aware shift planning, AI shift optimization, briefings in 5 languages |
| Tools & Assets | 🔧 | Fleet status, battery/health monitoring, vibration anomaly detection, predictive maintenance, low-carbon charging windows |
| Impact Reports | 📊 | Safety/carbon/productivity summaries, exportable reports, mapping to GHG Protocol, GRI, CIDB NCP 2030, ISO 45001 |
🔗 One Alert. Whole-Platform Ripple.
Civora's biggest differentiator isn't any single screen — it's that these aren't seven separate dashboards.
🌡️ Heat-Risk Alert Detected (MEP Corridor B) ripples instantly into:
| Destination | Effect |
|---|---|
| 🧭 Command Center | Priority intervention card appears instantly |
| 🛡️ Safety Register | Risk + explainability breakdown updates |
| ✅ Accountable Task | Dispatched & logged to supervisor |
| 📊 Impact Report | Evidence trail updated automatically |
One signal. Four systems updated. Zero manual re-entry.
🎯 Example: The Heat-Risk Scenario in Action
| Step | What Happens | Key Metric |
|---|---|---|
| 1️⃣ | Risk detected across temperature, ventilation, task intensity, exposure time & fatigue | — |
| 2️⃣ | Civora surfaces the alert with full transparency | 92% confidence · 18-min window |
| 3️⃣ | AI recommends cooling station + crew rotation + shorter work blocks | — |
| 4️⃣ | Supervisor reviews and clicks Approve | Human-in-the-loop ✅ |
| 5️⃣ | Decision logged to audit trail with timestamp | 14:30:16 |
| 6️⃣ | Outcome feeds Safety, Productivity & Carbon impact metrics | Linked automatically |
💬 "One click — and it's logged, assigned, and accountable."
⚙️ How We Built It
| Layer | Technology | Why It Matters |
|---|---|---|
| 🎨 UI | React 19 | Modern, fast, component-driven |
| 🧠 Language | TypeScript 6 (strict) | One shared domain model across all 7 workspaces |
| ⚡ Build | Vite 8 | Lightning-fast dev & optimized production builds |
| 📈 Visualization | Recharts 3 | Live telemetry, forecasts, carbon scenarios |
| 🎭 Icons | Lucide React | Clean, consistent iconography |
| 📴 Offline | Vite PWA + Workbox | Installable, works without connectivity |
| 🧪 Testing | Vitest + Testing Library | 7/7 tests passing on risk ranking & safety scoring |
| 💾 Persistence | Browser local storage | Alerts, tasks, projects, theme survive refresh |
🔬 Reliability by Design
To make the demo bulletproof for judging, we built a deterministic simulated telemetry layer updating every 4 seconds — no flaky third-party APIs, no network dependency, no surprises on stage.
| Quality Gate | Result |
|---|---|
| TypeScript strict check | ✅ Passed |
| Automated tests | ✅ 7/7 Passed |
| Production build | ✅ Passed |
| npm security audit | ✅ 0 Vulnerabilities |
| Route-level code splitting | ✅ Enabled |
| Offline PWA generation | ✅ Enabled |
🧗 Challenges We Ran Into
| Challenge | How We Solved It |
|---|---|
| 🧩 Avoiding a "pile of dashboards" | Designed every workspace around the same Sense → Understand → Recommend → Approve → Act → Verify loop, sharing one state model |
| 🤖 Making AI credible, not a black box | Built in confidence scores, factor-contribution breakdowns, data-freshness indicators, and mandatory human approval |
| 🔍 Being honest about what's real | Clearly disclosed deterministic demo telemetry everywhere, while architecting for real sensor/BIM/HR/procurement adapters |
🏆 Accomplishments We're Proud Of
| ✅ | Accomplishment |
|---|---|
| 🔗 | One integrated workflow connecting safety, sustainability, workforce, tools & reporting — not 7 disconnected screens |
| 🌡️ | Explainable heat-risk intervention with confidence scoring, time-to-threshold & accountable dispatch |
| 🗺️ | Interactive spatial digital twin with live telemetry layers & 2-hour risk forecasting |
| 🌱 | Carbon scenario modeling (baseline vs. AI-optimized) + real circular material matchmaking |
| 🗣️ | Multilingual briefings across 5 languages with comprehension tracking |
| 🔧 | Predictive maintenance derived from simulated tool vibration drift |
| 📴 | Fully installable, offline-capable PWA — persistent state, automated tests, clean build, 0 vulnerabilities |
📊 Live Outcome Snapshot (Demo Data)
| Metric | Value | Trend |
|---|---|---|
| 🛡️ Site Safety Score | 87 / 100 | ▲ +4.2% |
| 🌱 Carbon Avoided | −142 tCO₂e | vs. baseline |
| 📈 Productivity Index | 94% | ▲ +2.1% |
| 🔧 Fleet Uptime | 98.6% | ▲ +0.8% |
| 👷 Active Workforce | 142 / 168 | ▲ +2% |
🎬 One Decision. Three Outcomes.
| 🟢 SAFER | 🔵 LEANER | 🟠 LOWER-CARBON |
|---|---|---|
| Crew rotated before threshold | Tool serviced before failure | Material reused, not landfilled |
💬 "One decision. Three outcomes — safer crews, higher utilization, lower carbon."
📚 What We Learned
- Trust is half the innovation. A confident-looking AI score means nothing if a supervisor can't challenge it. Accountability, privacy, and visible data quality matter as much as the model itself.
- Sustainability becomes actionable when it's daily, not monthly. Translating carbon and waste into real-time site decisions and financial terms changes behavior — a quarterly spreadsheet doesn't.
- One shared domain model early = coherence later. Designing sites, zones, alerts, tasks, workers, tools, and materials as stable TypeScript contracts from day one made 7 workspaces feel like one product, not seven small ones.
🗺️ What's Next for Civora
Phase 1 — Pilot (6–8 weeks) 🧪
| Action | Goal |
|---|---|
| 🌡️ Connect real environmental sensors + limited BIM model | Ground-truth the risk models |
| 🔧 Integrate live tool/asset telemetry | Validate predictive maintenance |
| 📏 Validate alert precision & false-positive rates | Build supervisor trust |
| ⏱️ Measure response time, exposure minutes avoided, downtime | Prove ROI |
| 🗣️ Interview workers on privacy & briefing comprehension | Human-centered validation |
Phase 2 — Scale 📈
| Action | Goal |
|---|---|
| 🌱 Expand carbon & material-circularity workflows | Full sustainability loop |
| 🔐 Add enterprise identity & role-based access control | Enterprise-ready |
| 🧠 Prospectively validate safety models | Certifiable accuracy |
| 📋 Build audit & evidence-trail infrastructure | Assurance-ready reporting at scale |
🏗️ Build intelligently. Protect people. Waste less.
Civora helps construction teams see sooner, decide better, and prove what changed. 🌍
Built With
- css
- lucide-react
- pwa
- react
- recharts
- testing-library
- typescript
- vite
- vite-pwa
- vitest
- workbox

Log in or sign up for Devpost to join the conversation.