Back to Case Studies
Deep Loom · Success Story

Scaling Dutch Therapeutic
Science to the World

SilentHands had a method that worked and a delivery model that didn't scale: one coach, one room, one session at a time. We built the AI that runs a full session on any device. Investors backed the proof of concept before the product was finished.

AI & HealthcareComputer VisionAmsterdam to Worldwide
SilentHands - Scaling Dutch Therapeutic Science to the World
SilentHands
99.2%
Gesture Classification Accuracy
92%
Session Completion Rate
90%
Cost Reduction per Session
29
Gesture Classes Recognised
Milestone

The short version

A decade of peer-reviewed therapy, stuck behind a coach's calendar. We turned the method into an AI system that guides a full session, checks the hand movements on camera, and measures the result. Investors backed the proof of concept in April 2026. Full production continues from May.

The Problem

A Method That Works. A Delivery Model That Couldn't Scale.

Since 2007, Rogier Gielen has helped people resolve trauma, grief and anxiety with structured hand movements. Clients describe life-changing results in a single session. Peer-reviewed research backs it up.

Every session needed a certified coach, a room in Amsterdam, a booked slot and a client who could afford it. The people who needed it most couldn't get near it.

The therapy was never the limit. The delivery was.

Bottleneck

One coach, one client, one room at a time

Geography

If you weren't near Amsterdam, you were out

Cost

Session pricing excluded most of the world

Data

No measurement. Progress was self-reported

“I am so happy with my hand movement. It brings me back to who I essentially am. Since applying it to old traumas, I enter a peace and stillness that I have never felt so strongly before.”

Carolien, SilentHands Client

“It is almost a miracle. My sleep is completely fine. I make my hand movement in my mind and then I relax and fall asleep. I feel so much better now, so much more rested.”

Marius, SilentHands Client

“After the death of my partner, I was very sad for a long time. After discovering my hand movement, I was able to feel joy again.”

Anonymous, SilentHands Client
What changed

From one office in Amsterdam to any device, anywhere

We didn't replace the therapy. We removed everything standing between it and the people who need it: the coach's calendar, the location, the price.

A session that needed an Amsterdam address now runs from a phone, in any country, at a tenth of the cost. And for the first time, it produces data.

90% Less
Estimated cost per session with AI delivery
Any Device
No location, no scheduling, no coach required
Global
Dutch-built method, available worldwide
SilentHands coaching in action - the exact therapeutic process, now replicated autonomously

SilentHands coaching in action - the exact therapeutic process, now replicated autonomously

The Solution

How it works

From November 2025 an Amsterdam team turned two decades of coaching into software, with one rule: replicate the real session, don't approximate it.

The session follows the same five-stage arc a trained coach uses: collect the memory, process it, integrate, regulate, apply. That fixed structure is why results are consistent without a human in the loop.

On top sits real-time AI. The camera checks each hand movement as it happens. Facial analysis measures tension instead of asking. A voice avatar guides the session in the tone of a live coach.

🤲
Hand Movement Recognition

Camera-based tracking checks every movement live. 99.2% accuracy across 29 gestures, no coach needed.

😶
Tension Measurement

Facial analysis measures emotional tension objectively. Self-reporting is gone.

🎙️
Voice Coach

Warm, context-aware guidance in the pacing and language of a trained SilentHands coach.

🔁
Session Engine

Six AI modules run a 15-20 minute session end to end. 92% of users complete it.

🧭
Five-Stage Arc

The exact structure coaches use in person, so outcomes repeat at scale.

📊
Session Data

Movement quality, tension and completion scores from every session. A feedback loop that never existed before.

Built in phases, validated early
Phase 0
Proof of Concept

Protocol codified, functional PoC built, 3-5 user tests conducted, strong enough to attract external investment.

Phase 1
Data & Annotation

Secure infrastructure, annotation dictionary, 15 session recordings labelled for model training.

Phase 2
Conversational Engine

Voice pipeline, LLM integration, avatar narration system operational.

Phase 3-4
Observer + Production

Custom gesture classifier, full system integration, end-to-end testing and production deployment.

The Team
Daniel Matias·Deep Loom

Lead AI/ML Engineer - gesture recognition, LLM orchestration, multi-agent system, full AI stack

Sander Korf·SDKode

Lead Software Engineer, 13+ yrs - full-stack application, infrastructure, scalable architecture

Anca Luca·The Roots

Social Designer - human-centered UX, ethical design systems, impact-driven product experience

Rogier Gielen·SilentHands

Founder - therapeutic protocol, domain expertise, client network, and investor

Tech Stack
MediaPipeGPT-4oElevenLabsLangChainPythonTensorFlowOpenAI APIClaude CodeSupabase
The Result

Autonomous sessions. Measurable outcomes.

The proof of concept did what the brief asked: a full therapeutic session with no coach, no room and no appointment. Movement recognition runs at 99.2% across 29 gestures. 92% of sessions complete. Tension is measured, not guessed.

Cost per session drops 90%. The system is language-agnostic, so new markets don't mean a rebuild.

Investors and real users validated it in April 2026, before the product was finished. Rogier has already demoed it to SilentHands clients. Full production continues from May 2026.

99.2%
Gesture Classification Accuracy
92%
Session Completion Rate
90%
Cost Reduction per Session
15-20 min
Autonomous Session Length
6
LLM Modules in Multi-Agent System
29
Gesture Classes Recognised