← All projects
Founder-built public productFounder and lead developer

Arrival Atlas

Flight reliability analytics

Historical flight reliability analytics for comparing directional routes, airlines, seasons, and departure windows using official U.S. flight records.

Arrival Atlas route reliability dashboard showing flight performance comparisons.
Live product interface using published aggregate data.
Case brief

The product, the pressure, and the response.

01

Overview

Arrival Atlas is a descriptive decision-support product, not a prediction of a specific future flight. It makes a large official dataset practical to explore while keeping sample sizes and methodology visible.

02

Context

Official flight records are detailed and high-volume, which makes direct browser analysis slow and makes simple averages easy to misread. Travelers need route-specific comparisons that preserve sample size, season, airline, and departure-time context.

03

Solution

An offline DuckDB pipeline validates and aggregates official records into compact Parquet partitions. A Next.js application loads only the relevant aggregate data and uses ECharts to present route, airline, airport, and historical comparisons.

Exact role

Founder and lead developer

Product definition, data pipeline, application engineering, testing, release, and deployment.

01

Directional route comparison

Compare airlines serving a selected origin-to-destination route without treating the reverse direction as the same journey.

02

Time-aware filtering

Inspect performance by month and departure window instead of relying on a single all-time average.

03

Reliability measures

Review on-time and cancellation rates alongside median and tail-delay measures.

04

Airport and trend views

Explore airport-level context and historical patterns from the same published aggregate dataset.

05

Transparent evidence

Keep methodology, observation counts, and sample-size safeguards visible with each decision-support view.

Delivery approach

Built in reviewable steps.

  1. 01

    Inspect and validate

    Profile official flight files, normalize fields, and apply documented data-quality rules before aggregation.

  2. 02

    Aggregate offline

    Use DuckDB to calculate reusable route, airline, airport, month, and departure-window summaries.

  3. 03

    Publish compact data

    Write versioned Parquet outputs so the public application does not ship or scan raw flight rows.

  4. 04

    Deliver and verify

    Build the responsive product, test its data paths, and verify the released application across desktop and mobile.

Verified outcomes

Evidence, with its limits visible.

36,533,897official flight rows processed
4,556,244aggregate rows published
382airports represented
8,434directional routes represented
MITsource-code license
Technology
Next.jsDuckDBEChartsApache Parquet
Current status

Live public product with an MIT-licensed repository and a reproducible aggregate-data workflow.

Discuss a related build

Start with the problem, not a copied feature list.

Start an inquiry ↗