Perle provides a human‑centric data and evaluation platform that connects AI teams with vetted experts—doctors, lawyers, engineers, and other specialists—to generate high‑quality, domain‑specific training data and safety assessments. Its three products include WhisperMind, a multi‑modal dataset pipeline for embodied AI, and a suite of data services that handle expert hiring, annotation, quality control, and model‑focused evaluation, all delivered by a global payroll of experts to ensure consistent standards and traceable audit trails.
Funding
$9M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Frontier AI models often fail on complex, high-stakes tasks because they lack high-quality, expert-generated data, evaluations, and rubrics. Crowdsourced labeling cannot provide the domain expertise needed for safety-critical applications such as medical diagnosis, legal reasoning, or embodied robotics.
Solution
Perle supplies a human‑centric data and evaluation platform that connects AI teams with vetted experts—doctors, lawyers, linguists, engineers, and scientists—across 70 countries and 27 languages. The company offers three integrated products: WhisperMind, a real‑world, multi‑modal dataset pipeline for embodied AI that captures synchronized vision, depth, pose, force, and audio streams labeled by domain‑experienced operators; and a suite of data services that handle expert hiring, annotation, quality control, and model‑focused evaluation. Their workflow combines automated pre‑labeling, multi‑tier expert review, and continuous calibration metrics, ensuring that every artifact is verified before delivery. By keeping experts on payroll rather than a marketplace, Perle maintains consistent standards and traceable audit trails for safety, bias, and refusal calibration.
Target Audience
AI teams developing safety‑critical or embodied models—such as robotics labs, healthcare AI developers, legal tech firms, and large enterprises needing expert‑level data and evaluation—are the primary customers.
Features
- WhisperMind platform captures egocentric, time‑aligned RGB‑D, IMU, force‑torque, and spatial audio data from human operators performing tasks in real environments
- Model‑in‑the‑loop annotation where an on‑platform model provides initial labels that experts refine, accelerating dataset creation
- Three‑tier expert review (annotator, senior specialist, Perle data lead) with inter‑rater agreement tracking and golden‑set validation
- Automated quality gates (linters, regression checks, rubric‑backed probes) that filter out obvious errors before human review
- Evaluation network of 2,500 physicians, 800 attorneys, and specialists across two dozen disciplines for red‑team testing, bias audits, and refusal calibration
- End‑to‑end production pipeline covering data collection, annotation, validation, and delivery at scale (terabytes per day)
- Global pool of 15 K+ vetted experts on payroll, supporting 27 languages and domain‑specific expertise