Engineers Launch AI Tracker Covering 20,000 Gas Stations
Two engineers built The Gas Index, an AI-driven national gas-price tracker that fills gaps left by Google Maps by combining public data, crowdsourced photos, and automated calls. The system uses image OCR to read station price boards, integrates Google data for major chains, and deploys AI-generated robocalls to query more than 20,000 stations that are otherwise unindexed. Users create vehicle profiles so the site factors distance, fuel efficiency, and octane when recommending the cheapest fill-up. The project began as a beer-price tracker and evolved into a national tool that estimates savings at the pump, converts price changes into relatable purchases, and prioritizes coverage in independent and remote locations. Engineers report the stack includes an LLM assist, using Claude for parts of development and automation, plus TTS and call automation to harvest prices at scale.
What happened
Two engineers, Matt Cortland and Jon Fleming, launched The Gas Index, an AI-powered national gas-price tracker that supplements Google data and reaches more than 20,000 stations Google Maps does not index. The tool aggregates public datasets, chain-reported prices, user-submitted photos, and automated phone calls to produce per-station prices and personalized recommendations based on vehicle profiles.
Technical details
The system combines several automated components. Key elements include:
- •OCR on user-submitted pricing-board photos to extract posted per-gallon prices
- •Google Maps and chain APIs for large national brands
- •AI-assisted robocalls and TTS to query independent and remote stations not covered by Google
- •An LLM assist using Claude for code generation and orchestration of automation tasks
- •A calculator that factors vehicle fuel efficiency, octane, and driving distance into cost-per-fill recommendations
The pipeline ingests crowdsourced images and call results, normalizes timestamps and price formats, and resolves conflicts between sources by recency and source reliability. Robocalls are scripted and scheduled to minimize disruption and to gather real-time price data where no digital feed exists.
Context and significance
Google Maps reports prices for under half of U.S. gas stations, leaving many independently owned or remote pumps unobserved. That visibility gap is a practical data problem for consumers and for any application needing complete local price signals. The Gas Index demonstrates a pragmatic approach to data coverage: combine structured public feeds with human-verified photos and active probing via calls. It also illustrates a growing pattern of LLMs and automation being used to orchestrate nontrivial data-collection workflows rather than only consumer-facing chat features.
What to watch
Accuracy depends on timely user contributions and the quality of automated call responses; expect ongoing iterations on deduplication, fraud/resilience checks, and rate-limiting of calling. Operators will need to balance scale with legal and ethical constraints around automated calls and data scraping.
Scoring Rationale
The project is a practical, well-executed application of AI and automation that meaningfully improves coverage of a real-world dataset, but it does not change core model or infrastructure paradigms. It is useful for practitioners thinking about data augmentation and LLM-driven automation.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.
