Data Science undergraduate at Universitas Brawijaya.
I compete in national data science competitions: computer vision, time series, survey data and Indonesian NLP. I pick models on cross-validation that matches the test split, chronological for time data and grouped for grouped data.
Can future greenhouse humidity be predicted without leaking the future into training?
Team of three. I built the reproducible pipeline for our 4-model ensemble: 8-fold chronological cross-validation and time-based interpolation of sensor gaps.
9thof 37 teams on the private leaderboard, RMSE 3.91
Browser model440 trees17 features96.5% of full QWK
Presented to the judges in the national final, team of three.
03 · Survey modelling · Universitas Airlangga, 2026
DataQuest 5.0: Life Satisfaction
How far can survey answers predict a person’s life satisfaction?
Expected-value thresholds replaced argmax and added 0.06 QWK. A learning curve over five data sizes flattened, which pointed to missing survey questions rather than model capacity.
Finalistnational final, with a model that runs in the browser
Indonesian comment classificationACTION 2026
Macro F1, grouped 5-fold CV0.919
Comments23,297
Topics, multi-label9
Models in ensemble3
IndoBERTIndoBERT-largeIndoBERTweet
Validated with folds grouped by topic.
04 · NLP · ACTION, 2026
ACTION 2026: Public-Service Topics
Which public-service topics is an Indonesian social-media comment about?
An ensemble of IndoBERT base, IndoBERT large and IndoBERTweet, validated with topic-grouped 5-fold cross-validation.
0.919macro F1 in grouped cross-validation, team of three
About
About me
I am in my third year of Data Science at Universitas Brawijaya (GPA 3.53/4.00). Most of my competition work is done in teams of three, and I also assisted the Database lab course in 2026.
I am free for a full-time internship from January to July 2027.