↓ Skip to main content

What I learned from NeurIPS Open Polymer Prediction (Kaggle)

·33 words·1 min
Ismail Can Oguz
Author
Ismail Can Oguz
I build ML-accelerated atomistic workflows (equivariant GNNs + DFT) to discover catalysts and understand surfaces.

Our team ranked 219 / 2240 (bronze). Notes:

  • Start simple baselines early; add ensembling later.
  • Careful CV splits matter as much as model choice.
  • Good feature hygiene and ablations beat one-off leaderboard jumps.