Meet Madhu: Your AI Yoga Companion

Yoga has always been a deeply personal practice — passed from teacher to student, breath to breath, generation to generation. But what if technology could amplify that relationship without replacing it?

That's the idea behind Madhu, OorjaKull's AI yoga companion. Madhu is not a replacement for your Acharya. It's a tool designed to give you real-time, camera-based pose feedback when you're practising at home, travelling, or simply can't make it to a studio session.

How Does Madhu Work?

Madhu uses your device's camera — whether that's a laptop, tablet, or phone — to analyse your body position in real time. Using a lightweight pose-estimation model, it identifies 33 key skeletal landmarks on your body: shoulders, hips, knees, ankles, wrists and spine.

Once your pose is detected, Madhu compares your alignment against a reference model trained on thousands of correctly-performed asanas. If your knee is drifting inward in Virabhadrasana II, or your hips aren't square in Trikonasana, Madhu will gently flag the correction — both visually on screen and through voice guidance.

The entire analysis happens in under 16 milliseconds, meaning the feedback feels instantaneous. There's no lag, no delay — just smooth, responsive guidance as you move through your practice.

Adaptive Breathwork

Madhu doesn't just track asana. It also guides you through adaptive breathwork protocols. Whether you're practising Nadi Shodhana (alternate nostril breathing), the 4-7-8 technique for sleep, or a NSDR (Non-Sleep Deep Rest) session, Madhu dynamically paces the inhale-hold-exhale cycle based on your session history and comfort level.

Over time, Madhu learns your breath capacity and gently extends your holds and exhales — training your nervous system without you consciously tracking numbers.

Why Madhu Doesn't Replace a Teacher

This is the most important part. Madhu can catch alignment issues, pace your breath and generate adaptive sequences — but it cannot read the room. It doesn't know you had a bad day. It can't feel the energy of a group class. It won't notice the subtle emotional release that happens in a deep hip opener.

That's why OorjaKull pairs Madhu with certified human Acharyas. The AI handles precision and consistency. The teacher handles presence, intuition and the sacred space of yoga.

The Technology Stack

For those curious about the technical side: Madhu runs on a WebAssembly-optimised pose detection pipeline. It uses a MoveNet-based model for skeletal tracking, with custom post-processing for yoga-specific joint angles. The breathwork module uses Web Audio API for voice synthesis and timing.

All processing happens on-device — nothing is sent to a server. Your camera feed stays completely private.

Try It Yourself

Madhu is available now on the OorjaKull AI platform. Whether you're a beginner learning Sun Salutation or an advanced practitioner refining your Bakasana, Madhu adapts to your level and grows with you.

Ancient wisdom, modern intelligence — that's the OorjaKull way.