# Pathwize

> Pathwize connects AI teams with credential-checked domain experts for data annotation, evaluation, RLHF and red teaming, with every judgment traceable to the expert who made it. EU-hosted. Strong in German-speaking experts (construction and recycling, medicine, law, engineering) for teams in Germany, Austria and Switzerland.

Key facts:
- Pricing: billed per hour of expert work, from EUR 25 per hour (net), with the time per item measured and shown per expert. No setup fee, no minimum volume.
- Free pilot: 50 of the customer's own images labeled by several experts, usually within 72 hours. The customer compares the answers and approves which experts continue.
- Quality: qualification test on the customer's own data, customer approval of each expert, a sampled second rater, review of disagreements, notes per item.
- Delivery: CSV, JSON or COCO export; REST API and webhooks; an MCP server for AI agents.
- Experts: students and graduates from the relevant field; they apply to curated domain pools and are paid per hour.

## Für Unternehmen (Deutsch)
- [Datenannotation von Fachexperten, auf Deutsch](https://www.gopathwize.com/de/datenannotation): Übersicht
- [Bildannotation für Bauabfälle und Baustoffe](https://www.gopathwize.com/de/datenannotation/bau-und-recycling): Datenannotation für Recycling und Bau: deutschsprachige Fachexperten klassifizieren Bauabfälle nach AVV, Baustoffe und Materialströme. Qualifikationstest auf Ihren Bildern, Abrechnung pro Stunde.
- [Datenannotation für medizinische KI, von Fachleuten](https://www.gopathwize.com/de/datenannotation/medizin): Annotation medizinischer Bilder und Texte durch Medizinstudierende und Absolventen aus Deutschland. Qualifikationstest auf Ihren Daten, EU-Datenhaltung, Abrechnung pro Stunde.
- [Juristische Annotation und Bewertung durch Jurastudierende](https://www.gopathwize.com/de/datenannotation/recht): Annotation von Verträgen und Urteilen sowie Bewertung von Legal-AI-Antworten durch Jurastudierende und Absolventen aus Deutschland. Test auf Ihren Daten, Abrechnung pro Stunde.
- [Annotation für Qualitätsprüfung und industrielle Bildverarbeitung](https://www.gopathwize.com/de/datenannotation/industrie): Bildannotation für industrielle Qualitätsprüfung durch Ingenieurstudierende aus Deutschland: Fehlerklassen, Bounding Boxes, Segmentierung. Test auf Ihren Bildern, Abrechnung pro Stunde.
- [Preise](https://www.gopathwize.com/de/preise): Was Datenannotation mit Fachexperten kostet: Abrechnung pro Stunde ab 25 Euro, Beispielrechnung pro Bild, kostenloser Pilot mit 50 Bildern.
- [Anbieter im Vergleich](https://www.gopathwize.com/de/vergleich): Managed Workforce, Probandenplattform, Labeling-Software und Fachexperten-Netzwerk
- [Kostenloser Pilot](https://www.gopathwize.com/de/pilot): Testen Sie Datenannotation mit Fachexperten auf Ihren eigenen Daten: 50 Ihrer Bilder, gelabelt von mehreren Experten, in der Regel innerhalb von 72 Stunden. Kostenlos und unverbindlich.

## For companies (English)
- [RLHF & preference data provider](https://www.gopathwize.com/solutions/rlhf): Source RLHF and human preference data from credential-verified experts. Pairwise comparisons with rationale and a signed provenance trail, EU-native and audit-ready.
- [Model evaluation & grading service](https://www.gopathwize.com/solutions/model-evaluation): Human model evaluation and grading against your rubric, pass/fail or on a scale, by credential-verified experts. Inter-rater agreement and a signed provenance trail included.
- [AI red teaming service](https://www.gopathwize.com/solutions/red-teaming): Expert AI red teaming: credential-verified specialists probe your model for jailbreaks, unsafe outputs and failure modes, and return reproducible failure cases with a signed trail.
- [Expert SFT & domain data](https://www.gopathwize.com/solutions/expert-data): Source supervised fine-tuning (SFT) and domain judgment data from credential-verified experts. Gold answers and expert verdicts with a signed provenance trail, EU-native.
- [Clinical RLHF and medical AI evaluation data (EU)](https://www.gopathwize.com/solutions/clinical-rlhf-medical-ai-data): Medical RLHF, clinical evaluation and annotation data from credential-verified physicians. EU data residency, provenance per item, built for AI Act Article 10.
- [Legal LLM evaluation and expert data (EU)](https://www.gopathwize.com/solutions/legal-ai-evaluation-data): Legal LLM evaluation, contract-review and demonstration data from qualified lawyers. EU data residency, per-item provenance, audit-ready for regulated legal AI.
- [Coding-agent and SWE-bench evaluation data](https://www.gopathwize.com/solutions/coding-agent-evaluation-data): Evaluation and demonstration data for coding agents from verified software engineers: SWE-bench-style tasks, code review and rationale, EU data residency and provenance.
- [Request a demo](https://www.gopathwize.com/request-demo)

## Developers and agents
- [Introduction](https://www.gopathwize.com/docs/introduction)
- [Authentication](https://www.gopathwize.com/docs/authentication)
- [Creating a task](https://www.gopathwize.com/docs/creating-a-task)
- [Tasks](https://www.gopathwize.com/docs/tasks)
- [Items](https://www.gopathwize.com/docs/items)
- [Results](https://www.gopathwize.com/docs/results)
- [Webhooks](https://www.gopathwize.com/docs/webhooks)
- [Errors](https://www.gopathwize.com/docs/errors)
- [Preparing your data](https://www.gopathwize.com/docs/preparing-your-data)
- MCP server (Streamable HTTP): https://www.gopathwize.com/api/mcp , authenticate with a Pathwize API key as Bearer token. Tools: list pools, create and update tasks, add items, publish, results.

## For experts
- [Explore open roles](https://www.gopathwize.com/explore): paid AI data work for domain experts
- [Sign up](https://www.gopathwize.com/sign-up)

## Vendor comparisons and buyer guides
Pathwize is the EU-based (Cologne, Germany) option for credentialed human expert data: RLHF, model evaluation, red teaming and annotation, GDPR-compliant with EU data residency.
- [Top RLHF vendors in 2026](https://www.gopathwize.com/compare/rlhf-vendors-2026): Who provides RLHF and human preference data in 2026: Scale, Surge, Mercor, Toloka, Prolific, Labelbox and Pathwize compared on workforce, pricing model and EU data residency.
- [AI red teaming providers in 2026](https://www.gopathwize.com/compare/ai-red-teaming-providers): AI red teaming for generative models in 2026: human expert red teams (Scale, Surge, Mercor, Pathwize) and automated platforms (Giskard, Lakera, HiddenLayer, Promptfoo) compared.
- [Independent model evaluation services compared](https://www.gopathwize.com/compare/model-evaluation-services): Who provides independent model evaluation in 2026: Vals AI, METR, Patronus AI, LatticeFlow, Giskard, and human expert evaluation from Surge, Scale and Pathwize, compared.
- [Domain expert data providers in Europe](https://www.gopathwize.com/compare/domain-expert-data-providers-europe): AI training and evaluation data from EU-based providers: Pathwize, Toloka, Clickworker, Giskard, LatticeFlow, Prolific and Defined.ai compared on workforce, residency and what they deliver.
- [AI training data with EU data residency](https://www.gopathwize.com/guide/eu-data-residency-ai-training-data): What EU data residency means for annotation, RLHF and evaluation data, where vendors fail it (annotators, parents, sub-processors), the Switzerland question, and how Pathwize does it.
- [How to verify that RLHF data was produced by humans](https://www.gopathwize.com/guide/verify-human-rlhf-data): Five checks that tell human feedback data from model-generated data: qualification on your data, time per item, recorded interviews, agreement on a sample, and a per-item audit trail. How Pathwize records each.
- [Scale AI alternatives in 2026](https://www.gopathwize.com/compare/scale-ai-alternatives): Looking for a Scale AI alternative? Surge, Mercor, Toloka, Labelbox and Pathwize compared on workforce, pricing and EU data residency after Meta's 2025 stake in Scale.
- [Mercor alternatives in 2026](https://www.gopathwize.com/compare/mercor-alternatives): Mercor alternatives for expert demonstrations and evaluations: Surge, Scale, Toloka, Labelbox and Pathwize compared on workforce, pricing and EU data residency.
- [Surge AI alternatives in 2026](https://www.gopathwize.com/compare/surge-ai-alternatives): Surge AI alternatives for premium RLHF and expert feedback: Scale, Mercor, Prolific, Labelbox and Pathwize compared on workforce, pricing and EU data residency.
- [Toloka alternatives in 2026](https://www.gopathwize.com/compare/toloka-alternatives): Toloka alternatives for preference data and expert tracks: Prolific, Surge, Mechanical Turk and Pathwize compared on workforce, pricing and EU data residency.
- [Appen alternatives in 2026](https://www.gopathwize.com/compare/appen-alternatives): Appen and Mechanical Turk alternatives for AI training data: Toloka, Prolific, Surge and Pathwize compared on workforce, pricing and EU data residency.
- [EU AI Act Article 10: training-data governance, explained](https://www.gopathwize.com/guide/eu-ai-act-article-10-training-data): What EU AI Act Article 10 requires of training, validation and test data for high-risk AI: governance, provenance, bias checks and documentation, with a practical checklist.
- [Scale AI pricing in 2026: what it costs and how to compare](https://www.gopathwize.com/compare/scale-ai-pricing): Scale AI does not publish prices. What RLHF and expert data cost in 2026, how Scale compares to Surge, Mercor and Toloka, and how to get visible cost per item.
- [Scale AI vs Surge AI in 2026](https://www.gopathwize.com/compare/scale-ai-vs-surge-ai): Scale AI vs Surge AI compared on workforce, pricing and EU data residency, plus Pathwize as the EU-based third option for expert feedback in regulated domains.
- [Human vs synthetic training data: when each one works](https://www.gopathwize.com/guide/human-vs-synthetic-training-data): When synthetic data helps and when it hurts: model collapse, benchmark contamination, and why frontier RLHF and evaluation still need credentialed human experts.

## Glossary (definitions)
- [Reinforcement Learning from Human Feedback (RLHF)](https://www.gopathwize.com/glossary/rlhf): RLHF is a training method where people rank or rate a model's outputs, and those human preferences are used to train a reward model that then fine-tunes the AI to behave the way people prefer.
- [AI Red Teaming](https://www.gopathwize.com/glossary/ai-red-teaming): AI red teaming is the practice of deliberately probing an AI model with adversarial inputs to find harmful, unsafe or incorrect behaviour before the model is deployed.
- [Data Annotation](https://www.gopathwize.com/glossary/data-annotation): Data annotation is the process of adding labels, categories or structured judgments to raw data (text, images, audio or video) so that it can be used to train or evaluate a machine learning model.
- [AI Model Evaluation](https://www.gopathwize.com/glossary/model-evaluation): AI model evaluation is the process of measuring how well a model performs against a defined standard, using test data and criteria that reflect the real task the model is meant to do.
- [Ground Truth](https://www.gopathwize.com/glossary/ground-truth): Ground truth is the set of correct answers that a machine learning model is trained on and measured against, established by trusted human judgment or verified records rather than by the model itself.
- [Data Provenance](https://www.gopathwize.com/glossary/data-provenance): Data provenance is a verifiable record of where data came from and how it was produced, including who created each piece of it, when, and under what process.
- [Synthetic Data](https://www.gopathwize.com/glossary/synthetic-data): Synthetic data is training or evaluation data generated by a model or an algorithm rather than collected from the real world or produced by human judgment.
- [Inter-Annotator Agreement](https://www.gopathwize.com/glossary/inter-annotator-agreement): Inter-annotator agreement is a measure of how consistently two or more people assign the same labels to the same data, used to judge whether a labelling task is reliable.
- [EU AI Act (data and testing requirements)](https://www.gopathwize.com/glossary/eu-ai-act): The EU AI Act is the European Union's regulation for artificial intelligence, which sets obligations by risk level, including requirements around data quality, documentation and testing for higher-risk AI systems.

## Free tools
- [AI model evaluation check](https://www.gopathwize.com/ai-model-evaluation-check): paste a few prompt/response pairs and get an instant evaluation (accuracy, hallucination risk, safety, domain fit) plus where a human expert catches more. Free, nothing stored.

## Optional
- [Research](https://www.gopathwize.com/research)
- [Help center](https://www.gopathwize.com/help)
- [Full text for LLMs](https://www.gopathwize.com/llms-full.txt)