Geolocation & Real Estate over Brazil
summary
Describes building a geolocation service that fixes legacy "location ID" strings and geocodes six million listings using Scala and Akka Streams, a step-based pipeline, and a test harness called Columbus Monaco. Claims throughput limits (50 rps raised to 300 rps) and geocode latency (~800 ms) force caching, throttling and validation steps; precision categories like rooftop precision drive product metrics. A/B statistical tests and dashboards show rooftop hits rise and downstream features (recommendations) lift conversion by about 30%, enabling faster, data-driven iteration.
AI-written summary · about
chapters (37)
- 0:00Intro
- 2:05Brenno Costa
- 2:21Natto
- 2:50Christmas
- 3:09Ponta Negra
- 3:33Black Point
- 4:40Agenda
- 5:28Project Challenges
- 6:26Legacy
- 7:39Location ID
- 8:40Addresses
- 9:25Christopher Columbus
- 10:21End Point
- 10:45Staging Point
- 11:19Precision Field
- 12:17Ranged Inference
- 12:57Geometric Center
- 13:22Approximate
- 13:51GeoCode API
- 14:36Client Input
- 15:44Missing Misplaced Information
- 17:22Free Text
- 18:42Limitations
- 23:01Step
- 24:18Context
- 25:00Steps
- 27:05Magics
- 29:58Feedback
- 31:31Solving the problem
- 32:35Columbus Monaco
- 33:38Scala
- 37:10Java Script
- 37:29Comparison
- 38:27Statistics
- 39:55Frequency Graph
- 40:33Wrap up
- 42:22Where to find Brenno
related talks
23mDistributed System Development with ScalaLoci
38mCrunching data in real time for Fortune500 clients
31mToo Far Too Furious: Multi-Region Cloud Applications
47mMonolith to Microserv. at 20M+ requests per second: Latency and Scale Challenges
06mInterview
50mDistributed Tracing in the Wild
21mBuilding Distributed Platforms at Extreme Scale on Fast Changing Data
22m"How We Built a Rust-y Real-Time Public Transport Map"