KeepCool

Inspiration

We were inspired by the staggering economic and human cost of depression, specifically in Hong Kong, where it generates over HK$2.51 billion in annual healthcare costs. We identified two critical failures in the current system: it's largely reactive, only engaging during crises, and it's fragmented, with wearables, apps, smart homes, and clinicians operating in isolated silos. We founded KEEPCOOL to bridge these gaps, creating a proactive, integrated "Mental Health Angel" that enables early intervention and seamless collaboration.

What it does

KEEPCOOL is an AI-powered mental health ecosystem that acts as a digital guardian. It passively collects data from your voice, wearables (e.g., heart rate, typo rate, breathing rate), smart environment, and personal journaling. Our proprietary AI analyzes this data to generate a real-time "Wellness Score." Based on this score, KEEPCOOL delivers ambient interventions—like adjusting lighting or suggesting meditation—and, with user consent, can alert a therapist and book urgent appointments, creating a closed-loop of care between the individual and professionals.

How we built it

We built a secure, multi-layered platform:

Data Collection Layer: Secure APIs and integrations gather consented data from user inputs (micro-journaling, mood check-ins), wearables, and IoT devices (smart lights, speakers).
AI Analysis Engine: Our core is a proprietary algorithm that fuses these data streams to calculate a personalized Wellness Score and integrate it into a professional mental report.
Intervention Layer: This score triggers personalized actions, from simple suggestions to critical alerts sent to integrated healthcare professionals.

Challenges we ran into

Breaking Down Silos: Integrating disparate data from various wearables and smart home brands required building complex, reliable APIs and navigating different data protocols.
Privacy-Preserving Personalization: Our biggest challenge was designing an AI that delivers deeply personal insights without compromising privacy. We solved this through on-device processing and strictly anonymized analysis.
Clinical Validation: Ensuring our algorithms and interventions are grounded in clinical science, not just correlation, requires ongoing collaboration with mental health experts to avoid bias and ensure efficacy.
Wearable system connection: Different wearables have different systems. We are still experimenting with various applications; however, we lack the revenue for the smart home devices and system for testing.

Accomplishments that we're proud of

We have tried to develop a basic AI model for the chatbot. Even though the model is rough and still requires further refinement, it encourages us to improve the structure and be more prepared for function in the next version.

What we learned

We learned the immense value of ambient, passive data (voice, sleep patterns) in understanding mental states beyond self-reporting. We gained a deep appreciation for the necessity of human-in-the-loop design, where AI augments rather than replaces clinicians. Most importantly, we learned that in mental tech, trust is the product; every decision must be made through the lens of privacy and ethical data use.