Early detection of side effects through AI-supported video analysis

The customer

Likeminded, now nilo, provides simple and direct access to mental health support. The Berlin-based company helps organizations and their employees to strengthen their mental resilience and general well-being in the long term. The aim is to break down the stigma surrounding mental health and enable as many people as possible to improve their mental health in a self-determined way. To this end, nilo provides individually tailored offers via a user-friendly, multilingual platform - from 1:1 coaching sessions to group sessions and independent exercises.

The challenge

nilo offers a wide range of mental health support - from individual coaching sessions to group sessions and self-guided exercises. To help users choose from the wide range of content and formats, the company planned to develop a recommendation system that would match users with suitable offers. Merantix Momentum was commissioned to lay the foundations of the system along three main lines:

Modeling a digital twin along the dimensions of anxiety, depression and burnout
Development of a model for recommending suitable psychological experts
Training of a format recommender who selects the appropriate form of offer (1:1, group, self-help)

"The collaboration with Merantix Momentum helped us to lay the technical foundation for our recommendation system. The clear recommendations for action and thinking from both a user and data science perspective made the project very valuable."

Josu Garcia de Albizu Armentia

Co-Founder & CPTO, nilo

Solution

To lay this foundation, Merantix Momentum worked intensively with nilo for three weeks. In phase 1, we advised on the data concept: decision factors were identified in deep-dive workshops with psychologists and an information-rich data set was curated.

After collecting sufficient labeled data, phase 2 started: validation of hypotheses and model prototyping. Within two weeks, we performed statistical evaluations and developed an initial model of the digital twin that achieved 80% accuracy in the predictions. Finally, the project was handed over to nilo's engineering team - including recommendations for the further integration of machine learning.

Results & effects

We successfully launched nilo's AI journey in just 3 weeks with tangible results and actionable recommendations.

Based on our prototype, nilo can predict psychological expert assessments with over 80% accuracy (anxiety, depression, burnout)

The collaboration was an important step in nilo's development towards becoming a personalized, AI-supported provider of evidence-based mental health solutions.

‍