Toxic gas identification — 18th out of 238

Ranked 18th out of 238 (top 8%) in the Toxic Gas Identification challenge from Bertin Technologies on ENS Challenge Data.

Task & approach

The task was to predict 23 alarm levels from 13 sensor signals. I worked with tree ensembles on this tabular prediction problem, comparing modeling choices and evaluating predictions against the challenge objective.

This project complements my work on geometric algorithms with practical supervised learning: translating sensor data into a prediction problem, selecting a model family and assessing performance through a shared benchmark.