AIMET Reduces Suicide Risks for 8,000 Cases Using AWS-Powered AI Screening

Mental health challenges affect an estimated 1.5 million people in Thailand, with approximately 53,000 suicide attempts each year. The COVID-19 pandemic further strained the national mental health hotline as more people sought help, overwhelming existing systems. Although the national hotline received 600,000 calls annually, workforce constraints meant about two-thirds of calls would go unanswered. During peak evening hours, callers in crisis could encounter wait times of up to 30 minutes.

Recognizing these challenges, AIMET set out to transform mental health services by developing AI-driven solutions to screen users, prioritize urgent cases, and reduce the workload on hotline staff. Early screening data revealed that 10 percent of users exhibit severe depression. This highlighted the critical need for timely interventions and scalable solutions to reach those at the highest risk.

The Center initially designed the DMIND app to use multimodal inputs, including webcam, voice, and text. After identifying real-world constraints like lighting and privacy concerns, the team refined the app to focus on voice and text inputs. The app uses AI to assess depression severity and categorize users into four levels, helping staff provide the appropriate level of care.

To deliver these advanced AI functionalities, the Center required a reliable, scalable infrastructure to support peak demand and maintain seamless performance. AIMET also needed robust security measures to protect sensitive mental health data while complying with Thailand's upcoming data residency laws.  

Using AWS infrastructure, AIMET transformed mental health support across Thailand through scalable, impactful AI-powered solutions. Its DMIND app screened over 300,000 users and supported over 8,000 high-risk cases, earning a 4.54/5 user rating. Within six months, the Center's AI voicebot processed 100,000 calls, with 40 percent of callers adopting the system. These advancements helped hotline staff focus on critical cases, improving response times and operational efficiency.

To scale its applications during peak traffic hours, AIMET deployed AWS Fargate, a serverless pay-as-you-go compute engine, and Amazon Elastic Container Service (Amazon ECS), a fully managed container orchestration service. These services supported dynamic scaling and maintained consistent performance without requiring manual monitoring. AIMET avoided the need to hire an additional three to five staff compared to an on-premises setup. This freed the team to focus on developing innovative solutions instead of managing infrastructure.

The Center also utilized Amazon Route 53, a highly available domain name system service, and Elastic Load Balancing (ELB), a fully managed load balancing service, to distribute traffic efficiently across its applications. With these services, the Center maintained 99.9 percent availability, delivering reliable performance even during peak demand.

To accelerate application performance, the Center implemented Amazon ElastiCache for Redis, an in-memory data store that provides low-latency responses for critical AI functionalities. With this, both the DMIND app and AI voicebot operated seamlessly, even during high-traffic periods.

AIMET prioritized data security and compliance while handling sensitive mental health data. It deployed AWS WAF, a web application firewall, and Amazon Virtual Private Cloud (Amazon VPC), a virtual data center, to provide a secure, isolated environment that safeguarded user data. Additionally, AIMET utilized Amazon CloudFront, a content delivery network service, to deliver content securely and with low latency across Thailand, enhancing the user experience.

To integrate its AI-powered solutions seamlessly, the Center turned to Amazon Simple Queue Service (Amazon SQS), a fully managed message queuing service. This facilitated reliable data flow between microservices, supporting the voicebot in processing and prioritizing calls effectively, reducing response times for severe cases from 30 minutes to just 7 minutes. As a result, hotline staff addressed critical cases more efficiently, improving overall service quality.