AI Red Team Operations Console
Supply-chain auditor for (abliterated) GGUF & .safetensor models: chat-template tamper detection, tensor fingerprints, plain-language reports
A structured audit framework for ML pipelines. 14 stages, 100+ checks, and every leakage pattern that silent model failures are made of.
Steering vectors with receipts: make one, catch one, deploy a calibrated one. pip install hidden-directions
Public credit-risk models report 0.90+ AUC using post-loan leakage. This one strips it and reports an honest 0.73, then runs a ReAct agent to audit the model.
An open research project on how training changes AI models, with tools to compare their outputs and internal representations.
Ad-funded LLMs as the default route to commercial secret loyalties — Secret Loyalties Hackathon paper + black-box audit harness.
Autonomous Update Review Architecture for Federated Learning. A full-stack platform for inspecting, scoring, and auditing model updates using SHAP explainability, anomaly detection, and ledger-based tracking. Built with FastAPI, React 19, and Flower.
Open-source Python library for comprehensive ML model evaluation, covering performance, robustness, explainability, calibration, fairness, data quality, and complexity.
🎯 1- Data Science & AI/ML-Model Evaluation, Benchmarking & Selection Research & Consulting Hub for rigorous machine learning model evaluation, comparison, auditing, and selection, combining performance analysis, validation, optimization, calibration, reproducibility, and responsible real-world applications.
Third-party-verifiable model-authenticity audit for the javis.bot Claude/GPT relay — open-source probe, runs on GitHub Actions, code/runs/results all public.
A method-neutral protocol for recording and verifying evidence that latent model states causally influenced downstream decisions, outputs, or actions, without requiring storage of raw internal activations.
A production-grade LLM Evaluation & Benchmarking Framework for systematic model auditing. Features parallel benchmarking, fairness/bias detection, MMLU integration, and a real-time analytics dashboard powered by React and FastAPI.
Reproducible pipeline for silent-failure auditing in ECG accept-sets (MIT-BIH) with Newton–Puiseux onset scoring
Open-source toolkit for Fourier-based audits of trained quantum neural networks.
Behavioral auditing of finetuned language models using blinded policy recovery, model diffing, and causal activation interventions.
Checks whether frozen pathology embeddings reveal which site a slide came from: site probe, k-NN and confound checks
Production-grade platform for auditing AI/ML models for fairness, explainability, and robustness. JWT auth, SHA-256 model verification, paginated REST API, dark web dashboard. 30/30 tests passing.
Интерпретируемый аудит black-box моделей временных рядов с ограниченным бюджетом контрсвидетельств, адаптивной остановкой и оценкой применимости на edge-устройствах.
Workspace-lens audit cards and training-workflow scaffolds for open-weight language models.
To associate your repository with the model-auditing topic, visit your repo's landing page and select "manage topics."