The library

nutrition-science-foundations · When you need foundational knowledge of how macronutrients, micronutrients, and digestion work.

skillx_class_aware_c2__feature_engineering__SKILL · Preparing respondent-keyed data for difference-in-differences causal analysis.

ml-pipeline · Best for designing end-to-end ML infrastructure with experiment tracking and pipelines.

pandas-pro · When manipulating large datasets with grouped aggregations, merges, time series resampling, or memory optimization.

sql-patterns · When optimizing queries, designing schemas, or ensuring SQL injection prevention via parameterized queries.

postgres-pro · When tuning PostgreSQL for production performance via indexes, connection pooling, or replication strategy.

python-pro · When building production Python with strict typing, 80%+ test coverage, and async I/O optimization.

new-column · Schema evolution, adding new data fields, or incremental database schema changes

tables · Organizing structured data where tables provide clarity and queryability.

research-methods-psych · Psychology research where internal/external validity and replicability are non-negotiable.

rca-causal-inference · Incidents with rich quantitative data where you need defensible, reproducible, mathematical causal claims (not just narrative RCA), especially when distinguishing multiple confounding factors.

nnsight-remote-interpretability · Running the same interpretability code on GPT-2 locally and Llama-405B remotely without code changes, enabling scalable mechanistic interpretability research on massive models.

sparse-autoencoder-training · guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable f...

transformer-lens-interpretability · guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints a...

ray-data · Batch inference and preprocessing on 100GB+ datasets across multi-node clusters.

grpo-rl-training · Teaching specific output formats (XML, JSON) and verifiable tasks without preference pairs.

miles-rl-training · Training 1TB+ MoE models with speculative RL for 25%+ rollout speedup.

openrlhf-training · Scaling PPO/GRPO/RLOO/DPO training to 70B+ models with multi-node vLLM.

simpo-training · Quick preference optimization without reward model or RL infrastructure.

slime-rl-training · Research-grade RL training with flexible reward functions and algorithm variants.

torchforge-rl-training · PyTorch-native RL training with hardware acceleration and custom loss functions.

fine-tuning-with-trl · Multi-phase RLHF pipelines (SFT→Reward→PPO) where you control each alignment stage.

data-analysis-sci · When you have raw measurements and need to extract honest conclusions with proper error analysis.

verl-rl-training · Production math/reasoning tasks (GSM8K, MATH) where you need proven RL algorithms at scale.

experimental-design-sci · When you must isolate causal effects and design has constraints (ethics, cost, timescale).

scientific-method · When teaching or designing any empirical study and need a repeatable framework.

bayesian-methods · Incorporating prior domain knowledge into inference and updating beliefs with small sample sizes where frequentist confidence intervals fail

descriptive-statistics · Data exploration when hypothesis testing is premature and you need to understand raw data distribution first

inferential-statistics · Determining whether a sample observation generalizes to the population when descriptive stats alone insufficient

probability-theory · Building rigorous statistical reasoning from first principles when intuitive probability reasoning fails

evaluating-code-models · Comparing code model performance across standard benchmarks when new architecture or training method is evaluated

regression-modeling · Quantifying how variables influence outcomes when you need interpretable coefficients rather than pure prediction.

evaluating-llms-harness · Standardized model comparison using industry-standard benchmarks when you need reproducible academic metrics.

statistical-computing · Deriving confidence intervals and p-values when analytical formulas are unavailable or standard assumptions violated.

mlflow · Managing dozens of experiments with repeatable comparison and governance when you need audit trails.

experiment-tracking-swanlab · Real-time experiment monitoring during training when live curves beat post-hoc analysis.

tensorboard · When debugging deep learning models with real-time metric visualization and experiment comparison across runs.

weights-and-biases · When managing production ML experiments that need team collaboration, automatic metric logging, and hyperparameter optimization.

sql-patterns · When using sql-patterns is more effective than generic alternatives.

llamaindex · When using llamaindex is more effective than generic alternatives.

chroma · Open-source RAG prototyping where self-hosted storage is preferred over managed cloud.

gemini-deep-research · Multi-source synthesis tasks where systematic web search + AI analysis beats single-prompt research.

faiss · High-throughput vector search where metadata filtering is not required and GPU acceleration helps.

sentence-transformers · Text embedding when you need semantic vectors that are domain-tuned or when you want pure open-source.

guidance · Structured extraction when you need 100% format compliance and can define the grammar precisely.

instructor · Extraction tasks where you want automatic validation and retry without building your own harness.

outlines · Batch generation where you need guaranteed valid JSON/SQL/code and sampling speed is critical.

blip-2-vision-language · Zero-shot image understanding tasks where training data is unavailable and frozen backbones reduce compute.

clip · Quick zero-shot image classification and semantic search when training data unavailable and model is open-source.

ara-compiler · Ingest papers, code, experiments into falsifiable, agent-traversable knowledge package with cognitive + physical layers and provenance tracking.