Pinned Loading

1. clinical-model-validation-toolkit clinical-model-validation-toolkit Public

   Reproducible, synthetic-first validation harness for clinical prediction models: DeLong AUROC, calibration, cluster bootstrap, distribution-shift metrics, and a validation engine separating true sh…

   Python

2. clinical-rag-eval-harness clinical-rag-eval-harness Public

   Production-ready evaluation harness comparing Naive RAG, Advanced RAG, and GraphRAG on clinical notes — harm-encoding metrics (entity F1, concept resolution accuracy, false-confidence rate) with bo…

   Python

3. digital-twin-synthetic-controls digital-twin-synthetic-controls Public

   Credibility test of digital twins as synthetic control arms: does a survival model trained on one SUPPORT2 population predict a different population's real outcomes?

   Python

4. flash-mla-from-scratch flash-mla-from-scratch Public

   FlashAttention + Multi-Head Latent Attention, from scratch: Triton kernel, checkpointed PyTorch attention, MLA, and a reproduced roofline argument on real PTB-XL multimodal data.

   Python

5. medical-vlm-complexity-gradient medical-vlm-complexity-gradient Public

   Does architectural complexity help medical vision-language models? A three-tier complexity-gradient study on RSNA-sourced chest radiographs, benchmarked against documented Kaggle winners.

   Python

6. repurposing-domain-shift-cxr repurposing-domain-shift-cxr Public

   510(k)-style validation of adult-to-pediatric domain shift in a chest-radiograph pneumonia classifier, with DeLong/bootstrap/calibration stats and an interpretable KAN head

   Python