Cinder builds safety infrastructure for some of the most important AI and internet platforms in the world, including Midjourney, ElevenLabs, Synthesia and Black Forest Labs. We’re backed by Accel, Y Combinator, and Radical Ventures.
We’re building a network of specialists who test AI models for the ways they can cause real harm. When a customer project calls for your expertise, we bring you in on a paid, project-based contract.
We want to hear from:
Red teamers and jailbreakers with hands-on experience getting frontier models to break
Testers with experience across image, video, and audio generation, or in agentic and MCP-connected systems
Experts in specific domains, including child safety, CBRNE, violent extremism, NCII, cybersecurity, fraud, self-harm, psychology, and copyright
Native speakers who can test models in languages and cultural contexts beyond English
Projects vary in size and timing. Once you’re in the network, we’ll reach out when a project matches your background.
Design YAML-based adversarial prompt injection scenarios, test them against frontier models, and validate successful tasks that assess whether AI agents can be manipulated into violating safety policies.
Evaluate AI-generated finance, audit, insurance, compliance, and legal content by reviewing its accuracy and reasoning, identifying gaps, and providing structured feedback to improve model performance.
Evaluate AI-generated finance, audit, insurance, compliance, and legal content by reviewing its accuracy and reasoning, identifying gaps, and providing structured feedback to improve model performance.
Create realistic Excel tasks based on equity research modelling, solve each task to establish a correct answer, and write precise yes-or-no rubrics to evaluate AI performance on Research Associate work.
Evaluate Finnish-language GenAI content for accuracy and safety, improve training datasets, identify bias, and provide linguistic feedback that strengthens content moderation workflows and AI-generated responses.
