I compete on Kaggle as handudu. This portfolio collects my biomedical competition work and selected projects in computational biology and medical imaging.
Kaggle Competitions Expert (ranked 807 of 221,860; highest 540), with 7 silver and 4 bronze team competition medals; four of the silver medals are in biomedical imaging. Ranks are team ranks as listed on my Kaggle profile (October 2026). Full competition record
| Competition | Team rank | Medal | My role | Code |
|---|---|---|---|---|
| BYU - Locating Bacterial Flagellar Motors 2025 | 15 / 1136 | Silver | Prepared slice-level training data and local validation, trained YOLO-family detectors, and built multi-model fusion with 3D non-maximum suppression. | Notebook |
| Sartorius - Cell Instance Segmentation | 26 / 1505 | Silver | Trained MMDetection instance segmentation models and tuned class-specific score thresholds, minimum mask sizes and overlap removal. | Notebook |
| UW-Madison GI Tract Image Segmentation | 29 / 1548 | Silver | Built slice-stacking preprocessing and case-grouped validation folds, trained segmentation models, and ensembled them with post-processing. | — |
| RSNA Screening Mammography Breast Cancer Detection | 46 / 1687 | Silver | Built DICOM preprocessing and ROI cropping, trained image classifiers, and ensembled models with decision-threshold selection for the probabilistic F1 metric. | — |
The linked notebooks are public notebooks I published during these competitions; they are not presented as the teams' final submissions.
Seven additional competition results
| Competition | Team rank | Medal |
|---|---|---|
| WSDM Cup - Multilingual Chatbot Arena | 27 / 950 | Silver |
| Feedback Prize - Predicting Effective Arguments | 42 / 1557 | Silver |
| Feedback Prize - Evaluating Student Writing | 87 / 2058 | Silver |
| CMI - Detect Behavior with Sensor Data | 160 / 2657 | Bronze |
| Image Matching Challenge 2025 | 62 / 943 | Bronze |
| Eedi - Mining Misconceptions in Mathematics | 118 / 1446 | Bronze |
| PetFinder.my - Pawpularity Contest | 253 / 3537 | Bronze |
| Project | What it shows |
|---|---|
| Knee Diagnostic Labels Without a Gold Standard | Five public report-derived label sets for 4,349 knee MRI studies: aligning documented definitions for two sources resolves most of their disagreement, and simulations show how majority-vote references can favour models that share a labelling style. Results. |
| Image2Biomass: Validation Across Split Designs | Image regression of pasture biomass under four fold designs. Mean weighted R² falls from 0.68 with random folds to 0.58 with sampling-date folds and 0.49 with species folds; full results. |
| RNA Sequence and Structure Exploration | Exploratory analysis of RNA sequences and 3D coordinates; silver Notebook medal. |
| Coral Reef Object Detection with YOLOv5 and WBF | Ecological image detection with weighted box fusion; bronze Notebook medal. |
Other machine learning projects
| Project | What it shows |
|---|---|
| Time Series Validation for Tactical Allocation | Rolling-window validation and hyperparameter selection for market prediction; bronze Notebook medal. |
| Multilingual Preference Prediction with Gemma and LoRA | LoRA adapter inference with swapped-response augmentation; bronze Notebook medal. |
| Mathematical Reasoning with Code Execution | Annotated study of a public code-execution solution. |
| Tabular Fraud Classification with CatBoost | CatBoost baseline with stratified cross-validation. |
Notebook medals are awarded for public notebooks and are separate from the competition results above.
Code is archived as published on Kaggle; data and weights are not included.