Overview
What this challenge is about.
You receive a curated public multi-modal MRI brain-tumor cohort (~600 patients, T1/T1c/T2/FLAIR with whole-tumor / tumor-core / enhancing-tumor masks). Train all three architectures with comparable budgets. Evaluate Dice + Hausdorff-95 per sub-region on a held-out test split. Measure inference throughput in patients-per-hour on a single L4. Discuss the accuracy/throughput/training-cost trade-off. Deliver a 4-page memo for the clinical-AI team.
The Brief
What you'll do, and what you'll demonstrate.
Pick the best segmentation architecture for multi-modal brain-tumor MRI on Dice + Hausdorff + L4 inference throughput.
Earning criteria — what you'll demonstrate
- Apply standard medical-imaging segmentation architectures end-to-end
- Use Dice + Hausdorff-95 correctly and report per-sub-region performance
- Measure inference throughput on realistic GPU hardware
- Recommend a segmentation architecture under accuracy + throughput constraints
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Computer Vision Engineer
Architecture bake-offs with both clinical metrics and GPU-throughput reporting are the CV-engineer's headline portfolio piece at radiology-AI startups.
This challenge sharpens
- medical-imaging
- segmentation
- convolutional-neural-networks
ML Researcher
Fair multi-architecture comparison on a real medical-imaging benchmark is exactly the kind of focused study ML-research hiring loops grade.
This challenge sharpens
- segmentation
- benchmarking
- model-evaluation
MLOps Engineer
Reasoning about inference throughput per GPU directly bridges to MLOps work on serving medical-imaging models at scale.
This challenge sharpens
- benchmarking
- model-evaluation
- pytorch