How to win an AI Hackathon, without using AI
summary
Winning entry shows that simple engineering beats fancy models: Jumping Rivers ingests 5–8 GB of Azure CSVs and 15‑minute time series for DMAs, then uses R, Shiny, tidyverse, ggplot2 to assemble a single-page app in half a day that consolidates flow, pipe type/size, reported leaks and weather. The real win is operational: reduced engineer prep time to 10–15 minutes (from two hours), exposing that leakage estimates are noisy and the target is volume lost not leak count, and the app goes into production and wins competitions — won four hackathons without ML.
AI-written summary · about
chapters (6)
- 0:00Start
- 6:08Problem set up
- 10:57How do you estimate
- 12:31What do you need to
- 13:10What data does an engineer need?
- 14:05Solution: Reduce engineer time!
related talks
58mImproving disaster response with machine learning
28mUsing Python in a Data Hackathon
15mFast-tracking your AI Career with Kaggle
23mBig data and big insights
26mHow to develop useful and unbiased AI
34mForget about AI and do mathematical modeling instead!
20mUsing R, Python, and Cloud Infrastructure to Battle Aquatic Invasive Species
19mNow in AI - How We Got Here