Organization: The Trevor Project

Time-to-launch: 6 months (March 2021)

Collaborators: 40+ (ML Engineers, UXD, UXR, Software Engineers, Data Scientists, Product Managers, Internal Education & Clinical Services)

Problem
With the ambitious aim to serve 1.8 million youth, The Trevor Project recognized a need to expand its counselor base to manage crisis interventions effectively. However, the counselor training process faced limitations due to its manual and resource-intensive nature, hampering the scaling of the counselor workforce. The challenge lay in enhancing the training capacity without compromising the quality and efficacy of the training, ensuring counselors are prepared to navigate real-life crisis scenarios involving LGBTQ youth.

Solution
A machine learning-powered platform that automates and amplifies the counselor training process. CCS immerses trainees in realistic and relevant conversations with a simulator, programmed to emulate an LGBTQ youth in a crisis scenario. This not only ensures that the counselors are proficiently trained but also allows for the scaling of the training cohort, thereby increasing the number of graduating counselors.

Key Metrics

• Staff time saved per roleplay

• Increase in counselors trained YoY

• Realism & relevancy conversational scores

Key Features

• NLP conversational model

• Simulated crisis conversations

• Realism & relevance in training

• Integration into chat & text platform

• In-App survey for continuous improvement

Kendra’s Role

• Idea validation & need identification

• Product strategy & roadmapping

• Model & learning objective requirements engineering

• Integration & development

• UAT

• Training & team enablement

• Release planning

• Post-launch optimization

• Impact analysis

Technologies Used

• Google Cloud Platform

• GPT3 (fine-tuned on internal data)

• BigQuery

• Salesforce