kaggle-tools

kaggle-audio-bioacoustics

skills/kaggle-audio-bioacoustics/SKILL.md

Use when working on Kaggle audio/bioacoustics competitions like BirdCLEF — row_id parsing, pretrained bioacoustic backbones (Perch, BirdNET), per-species probes, temporal models, mel spectrograms, CV/LB gap, rare and zero-sample species, CPU inference budget

kaggle-code-competitions

skills/kaggle-code-competitions/SKILL.md

Use when working on any Kaggle code competition — notebook submission, data paths, offline packages, GPU/accelerator choice, kernel push workflow, time limits, or debugging "Submission Scoring Error", TIMEOUT and NO_SCORE

kaggle-competition-playbook

skills/kaggle-competition-playbook/SKILL.md

Use when starting, planning or steering any Kaggle competition — reading the metric, choosing CV, deciding which experiment to run next, interpreting CV/LB gaps, ensembling, and managing submission/GPU quota

kaggle-leaderboard-analysis

skills/kaggle-leaderboard-analysis/SKILL.md

Compare Kaggle public and private team positions, or plot an account's competition history as top percentages by year with readable paginated figures. Use for leaderboard shakeups, finishing positions, and competition progress reports.

kaggle-llm-finetuning

skills/kaggle-llm-finetuning/SKILL.md

Use when fine-tuning LLMs (LoRA/QLoRA, SFT, GRPO) inside Kaggle competitions — TRL/PEFT setup, packing, completion-only loss, data quality, adapter submission format, GPU quota and 12 h kernel limits

kaggle-llm-reasoning

skills/kaggle-llm-reasoning/SKILL.md

Use when working on Kaggle LLM reasoning/inference competitions like AIMO — model choice, vLLM on H100, tool-integrated reasoning (TIR), self-consistency voting, answer extraction, inference-server API, time budget, offline tokenizer files

kaggle-llm-serving

skills/kaggle-llm-serving/SKILL.md

Use when an LLM inference notebook on Kaggle GPUs (4x L4, 2x T4, single 24 GB cards) is throughput-bound — tokens/s too low for the runtime limit, choosing tensor vs data parallel, fitting a 20-30B model on one 24 GB card, KV-cache or Mamba/GDN state limiting concurrency, MTP/speculative decoding depth, int4/fp8 weights or KV, vLLM patch stacks (HyperQwen, Merlin), or benchmarking serving configs on RunPod before spending Kaggle quota

kaggle-medical-imaging

skills/kaggle-medical-imaging/SKILL.md

Plan and improve Kaggle MRI or CT classification pipelines with weak report labels, anatomical crops, multi-series inputs, teacher supervision, and site-aware validation. Use for medical-imaging competition experiments and solution reviews, not clinical diagnosis.

kaggle-tabular

skills/kaggle-tabular/SKILL.md

Use when working on Kaggle tabular, time-series or geospatial regression/classification — GBDT stacks (LightGBM/CatBoost/XGBoost), tabular foundation models (TabICL/TabPFN), spatial/group CV, feature ablations, stacking and hill-climbing, Optuna, local training + Kaggle inference