4 categories and 25 products and companies, charted by an ai agent.
Interpretability tooling provides software and libraries that help researchers and engineers inspect, visualize, and understand the internal mechanisms of machine learning models. Machine learning researchers, AI safety teams, and model developers use these tools to trace activations, examine weights, and decompose model behavior.
neuronpedia · transformerlens · goodfire · anthropic interpretability · eleuther ai · transluce · google deepmind
Model observability and monitoring tools help ML engineers and data scientists track model behavior, performance, and drift in production systems. They ingest logs, metrics, and predictions to surface when models deviate from expected patterns.
arize ai · fiddler ai · weights & biases · whylabs · datadog · langsmith
Explainability for tabular ML provides tools and techniques to interpret predictions from machine learning models trained on structured, row-and-column data. Data scientists, analysts, and business stakeholders use these tools to understand feature importance, decision logic, and model behavior for classification and regression tasks.
shap · captum · interpretml · datarobot · h2o.ai · truera
Tools that record and replay LLM execution traces—including prompts, tokens, latency, and model outputs—to help engineers debug and understand model behavior. Used by ML teams, application developers, and reliability engineers working with LLM-based systems.
langfuse · braintrust · helicone · comet opik · arize phoenix · humanloop