About the role
At Sword, we’re building AI to heal billions and unlock humanity’s full potential. In doing so, we’re pioneering AI Care, a fundamentally new approach to healthcare built for medical reasoning, safety, and real-time treatment, not generic technology applied after the fact. As both a clinical-centric frontier AI lab and an applied AI platform, Sword is reimagining how care is delivered at scale, removing traditional barriers like appointments, waiting rooms, and stigma so more people can access the care they need—and ultimately get back to lives lived in full.
Since 2020, Sword has expanded across physical therapy, women’s health, cardiometabolic, and mental health, and is now moving beyond the session to a fully AI-native, 24/7 care program that brings physical activity, therapeutic exercise, psychotherapy, nutrition, and behavior change into one connected experience. More than 700,000 members across three continents have completed over 10 million AI sessions, helping 1,000+ enterprise clients avoid more than $1 billion in unnecessary healthcare costs. Backed by 42 clinical studies, 44+ patents, and more than $500 million raised from leading investors including Khosla Ventures, General Catalyst, and Founders Fund, Sword is defining a new standard for healthcare.
Sword's AI models need to understand exercise quality the way a physical therapist does – not just whether a movement happened, but whether it was performed correctly, safely, and with adequate range of motion. Teaching an AI that requires human clinical judgment at scale.
Your role is to watch recorded exercise sessions and provide structured clinical assessments that become the ground truth to evaluate and train our AI. This is high-volume, repetitive work – you will assess many videos per day using consistent criteria. Quality and consistency across high volume is what makes the data valuable.
The Data & Tooling team produces the labeled datasets that power Sword's AI models. We build the training and evaluation data foundation that enables our AI to reason about human movement with clinical accuracy.
What you’ll be doing
Watching exercise session recordings and scoring movement quality using Sword's clinical assessment framework – covering injury risk, movement effectiveness, range of motion, technique, and overall execution
Providing rep-level assessments, session-level summaries, and structured written feedback
Reviewing and quality-checking assessments produced by other annotators on the team
Flagging ambiguous cases, unusual presentations, and content quality issues
Participating in calibration sessions with other annotators to maintain consistent scoring standards across the team
Working full-time (40 hours/week), remotely, using Sword’s web-based annotation tools. Part-time option is also available (min 20 hours/week)
What you need to have
Qualified physical therapist – active or recently practicing
Confident clinical judgment across a broad range of musculoskeletal physical exercises
Comfortable with high-volume, repetitive assessment work – this is the nature of the role
Strong attention to detail and ability to apply scoring criteria consistently
Fluency in English (all guidelines and feedback are in English)
Reliable internet connection and a desktop or laptop computer
Comfortable with web-based tools (full training provided)
