Description
Perit operates a specialized data marketplace designed to supply high-quality training data for artificial intelligence and machine learning applications. The platform offers access to 29 distinct datasets spanning 10 categories, including extensive collections of speech in 122 languages, egocentric and general video, production codebases, agentic reinforcement learning environments, field documents, and clinical cohorts. Each dataset listing provides crucial details such as volume, formats, licensing status, and availability, ensuring transparency for users.
Read More
The marketplace distinguishes itself by offering "data the internet doesn't have," emphasizing expert-led collection and curation. This includes two-speaker natural conversations, inbound support calls, simulated speech, long-form podcast conversations, and conversational speech in various Indic languages and English. For video, Perit provides first-person household task videos, licensed skilled activity videos, and a broad library across 75 content categories. Code and agentic RL datasets feature long-horizon agent tasks, production code excerpts from logistics and e-commerce, online-gaming platform backends, and full social-media organization data.
Perit also offers image and document datasets, including field photos, scanned personal and financial documents for OCR, and custom image annotation. Specialized data types like CAD & 3D projects, consumer pitch decks, enterprise workflow data (Slack, email, tickets), and de-identified clinical cohorts are available. The platform supports various modalities, including speech and audio, video, image, text, and embodied data for robotics teams, captured using diverse hardware like action cameras, LiDAR, Android phones, and custom rigs.
Perit's operational model is built on the principle that the person performing a task is the ideal individual to record, transcribe, and grade the data. Operated by Zyno AI Inc. and backed by Y Combinator, Perit maintains independence, ensuring that customer data does not co-mingle with competitors' held-out data. The company provides transparent pricing based on scope, locale scarcity, annotation depth, turnaround, and exclusivity, with quotes delivered within 48 hours. It also offers pilot batches for testing specifications and adheres to strict quality assurance protocols, with every batch shipping a detailed quality report.
- CategoryMachine Learning & Data Science
- Pricing ModelContact sales
- DeploymentCloud
- PlatformsWeb
Perit's Core Features
29 Licensed Datasets Across 10 Categories
Speech Data in 122 Languages
Egocentric and General Video Datasets
Production Codebases and Agentic RL Environments
Clinical and Enterprise Data Collections
Custom Data Collection and Annotation Services
Transparent Pricing with 48-Hour Quote Delivery
Pilot Batch Option for Specification Testing
Detailed Quality Reports with Every Data Batch
Perpetual, Non-Exclusive Data Licenses by Default
Contributor Consent Management and Data Security
SOC 2, ISO 27001, ISO 27701, TPN Gold Certified
GDPR, CCPA, and HIPAA Compliant Data Handling
Data Storage Pinned to Specific Geographic Regions
Physical Intelligence Data for Robotics Teams
How to use Perit?
Submit Brief: Provide a failure example (clip, photo, paragraph, task description) to initiate a quote.
Receive Quote: Get a written spec and quote within 48 hours, detailing quantity, locales, annotation depth, and contributor rate.
Review Sample: Receive a sample episode in your desired format within five working days.
Pilot Batch: Obtain a pilot batch with its QA report within two weeks to verify the specification.
Volume Delivery: Begin weekly data delivery cadence once the pilot batch clears quality checks.
License Data: Utilize licensed datasets for internal model development and evaluation under agreed terms.
Perit's Use Cases
- AI Model Training
- Speech Recognition Development
- Computer Vision Research
- Robotics Training
- Codebase Analysis
- Document Understanding
- Clinical AI Development
- Enterprise Workflow Automation
- Custom Data Annotation
- Agentic RL Environment Creation






