back to projects

TrustMyETA: Routing ETA Uncertainty Quantification

2025 / Berlin, Germany

Monte Carlo Sim, Geospatial Analysis, React, TypeScript

TrustMyETA: Routing ETA Uncertainty Quantification

A route-planning prototype that predicts arrival times as a probability range instead of a single number. Each route gets a Trust Score, computed from weather, traffic, incident data, and route geometry, to make clear how reliable the estimate actually is.

View Live Demo
1000+
Monte-Carlo Simulations
Uncertainty Quantification
<100ms
Latency
app
DEMO

// functional summary

TrustMyETA is a travel-time prediction system that calculates a Trust Score for any given route. Instead of a single ETA, it returns a probabilistic arrival-time range based on risk factors.

Figure above: dashboard view, risk analysis and trust score.

Core Logic

Uses Monte Carlo simulation (1000 iterations) to model travel variability.

Input Vectors

Analyzes route geometry (tortuosity), real-time weather (snow, visibility), and traffic patterns.

output

A normalized reliability score (0–100%), P95 worst-case arrival time, and categorized risk factors (safe / caution / high).

// system architecture

graph TD User[User] -->|Request ETA| Frontend[React Web App] Frontend -->|POST /api/analyze| Backend[Node.js / Fastify API] subgraph "Trust Engine (Backend)" Backend -->|Get Route| Routing[OSRM / Routing Service] Backend -->|Get Weather| Weather[Open-Meteo API] Backend -->|Get Traffic| Traffic[TomTom API / Mock Service] Routing -->|Route Geometry | Complexity[Complexity Analyzer] Weather -->|Forecast| Normalizer[Risk Normalizer] Traffic -->|Congestion| Normalizer Complexity --> RiskModel[Risk Assessment Model] Normalizer --> RiskModel RiskModel -->|Variance Factors| Simulator[Monte Carlo Simulator] Simulator -->|1000 Iterations| Confidence[Confidence Score & Range] Confidence --> Report[Reliability Report] RiskModel --> Report end Report -->|JSON| Frontend

// technical stack & skills

The project demonstrates proficiency in modern frontend engineering, geospatial analysis, and system architecture.

React 19 + Vite
High-performance build tooling and modern React patterns.
TypeScript
Strict typing for robust domain modeling.
Tailwind CSS
Utility-first, responsive design architecture.
Visualization
Complex data plotting with Recharts & Leaflet.

Backend & data science concepts

ArchitectureNode.js / Fastify (REST API) for high-throughput request handling.
ValidationZod schema validation enforces strict, typesafe API contracts.
AlgorithmsCustom Monte Carlo statistical modeling and standard-deviation analysis in pure TypeScript.

// technical challenge

One of the detailed challenges was efficiently running 1000+ Monte Carlo iterations in real-time on the client side without blocking the main thread, solved by optimizing the statistical modeling algorithms in TypeScript.

// key features

01

Probabilistic Estimation

Uses Monte Carlo simulations (1000 iter.) to go beyond single-point estimates, providing full probability distributions for arrival times.

02

Confidence Metrics

Calculates a unique 'Trust Score' (normalized 0-100%) and P95 worst-case scenarios to help users make safer travel decisions.

03

Multi-Input Analysis

Synthesizes route tortuosity, real-time weather data (snow/visibility), and historic traffic patterns into a single reliability metric.