Data Labeler

About into3.ai

into3.ai is India's first Learning Infrastructure Platform. Here, we treat learning very differently everything is science-backed. We monitor students' biomarkers in real-time facial expressions, cardiovascular parameters (SpO2, blood pressure, heart rate) and adapt the learning experience based on these signals. The result: learning isn't surface-level but relies on body markers that ensure actual knowledge transfer. Being AI-first, learning on into3 is hyper-personalised and unmatched in depth.

Deepscience Cognitech AI Labs Pvt Ltd | Noida | into3.ai

Why This Role Exists

You'll annotate student engagement data for into3's Real-Time Adaptive Engine. Watch student study session videos and label engagement states engaged, confused, distracted, bored across multiple signal channels (facial expressions, posture, gaze, etc.). Your labels directly train the AI model.

What You'll Actually Do (Not Corporate Fluff)

Week 1:

Learn the RTAE annotation schema: The 7 signal channels and what to label in each. Training sessions with the ML Engineer.

Practice labeling on 20-30 sample videos until you hit the inter-annotator agreement threshold with Data Labeler #2

Set up labeling tools (CVAT or Label Studio) and get comfortable with the workflow: watch annotate submit next

First Month:

Label student video data daily target: [ML Engineer sets daily targets based on pipeline needs]

Annotate engagement states: Highly engaged, moderately engaged, neutral, confused, distracted, disengaged with timestamps

Annotate signal-specific features: "Student frowned at 0:12" (facial), "Posture slumped at 0:45" (posture), "Looked at phone at 1:03" (gaze)

Weekly inter-annotator agreement checks with Labeler #2: Both of you label the same 10 videos independently, ML Engineer compares consistency

Flag edge cases: "Can't tell if confused or thinking" or "Camera angle makes facial expression unclear" these help the ML Engineer refine the schema

Ongoing:

Continuous labeling as more student data comes in post-launch the pipeline grows as into3 gets more users

Schema updates: As RTAE evolves, new signals may be added you'll learn and adapt

The Person We're Looking For

Someone patient, careful, and consistent. This isn't glamorous work. You watch videos of students studying and annotate what you see. For hours. Every day. The quality of your labels directly determines whether RTAE can tell if a student is confused and if it can't, the whole adaptive learning promise falls apart.

You can read human expressions and body language. You notice when someone's posture changes from attentive to slumped. You can tell the difference between a "thinking" frown and a "confused" frown. Not everyone can it's a real skill.

Must Have

Graduate in any discipline we train you on the annotation schema

Strong attention to detail and tolerance for repetitive work this is annotation, not analysis

Can recognize and categorize facial expressions, posture changes, gaze direction from video

Basic computer skills: Browser-based annotation tools, spreadsheets for tracking

Reliable internet connection for video streaming and annotation tool access

Punctual and consistent: Daily labeling targets must be met the ML training pipeline depends on your output

Bonus Points

Psychology, education, or child development background helps with understanding engagement signals

Experience with CVAT, Label Studio, Labelbox, or any annotation tool

Previous data labeling or annotation experience in any domain

This Role is NOT For You If...

You get bored doing the same type of work for hours. Labeling is repetitive by nature.

You're looking for a role where you can be "creative" or "strategic." This is detail work. Important detail work, but detail work.

You have inconsistent work habits some days 8 hours, some days 2 hours. We need predictable daily output.

You have concerns about watching student videos note: all videos are anonymized, and you'll sign DPDP compliance agreements. Student privacy is non-negotiable.