SafeStep

SafeStep tracking a patient taking their daily meals
SafeStep tracking a patient taking medication
Caregiver POV logging the patient's activity location
Patient navigating through simplified AI

Inspiration

The relationship between a caregiver and a care recipient often devolves into a cycle of nagging, checking, and anxiety. "Did you eat?" "Did you take your pills?" "Where are you?" These constant transactional questions erode the dignity of the senior and exhaust the patience of the caregiver.

We realized that to restore joy and meaning to this relationship, we needed to remove the friction of "surveillance." SafeStep was born from a desire to let technology handle the safety checks so that the humans can get back to simply loving each other. We aim to replace anxiety with assurance, allowing caregivers to provide care that is respectful of the senior's desire for independence.

What it does

SafeStep is a "Silent Guardian" platform that automates the friction points of caregiving, fostering a relationship built on trust rather than control.

Restoring Autonomy (Navigation & Safety): Seniors often stop going out because they fear getting lost or falling. Our Navigation & Fall Detection features allow them to move freely and independently, knowing help is automatic if needed.
Eliminating the "Did You?" Arguments (Activity Tracking): Instead of the caregiver constantly asking "Did you eat?", SafeStep uses Computer Vision to silently verify these tasks. The caregiver gets a green light on their dashboard, eliminating the need to micromanage.
Empowering Conversations (Data Summaries): By generating Medical Summaries automatically, we shift the doctor's visit from a stressful memory test to a meaningful conversation.
Non-Intrusive Support (Voice Reminders): Automated voice reminders provide gentle nudges for medication and meals, preserving the senior’s dignity by allowing them to self-correct before a caregiver has to intervene.

How we built it

We designed SafeStep to be an invisible layer of support, using a robust full-stack architecture that prioritizes real-time connection.

The Empathy Interface: Built with React 18 and Tailwind CSS, we designed two distinct dashboards. The Senior's view is simplified (high contrast), while the Caregiver's view is data-rich.
The "Silent" Backend: Using MongoDB Atlas, we sync activity logs in real-time. This immediacy is crucial—if a caregiver sees old data, they might panic unnecessarily.
Respectful Monitoring (AI/ML): We utilized YOLOv8 and MediaPipe for pose estimation. We calculate the net acceleration vector to determine a fall event:

$$A_{net} = \sqrt{a_x^2 + a_y^2 + a_z^2}$$

If the acceleration spikes and the pose remains horizontal, the system flags a "High Confidence" event where $ P(fall) > 0.8 $. This ensures the caregiver doesn't need to watch a live camera feed 24/7, preserving the senior's privacy.

Bridging the Gap: The Twilio API handles the "nudge" logic. It automates text updates (e.g., "Dad just took his medication"), transforming the caregiver's phone from a source of anxiety into a stream of reassurance.

Challenges we ran into

Balancing safety with privacy was our biggest technical and ethical challenge. We had to calibrate the Computer Vision (Python/YOLO) to be accurate enough to detect a fall without triggering false alarms that would erode trust. Connecting these heavy Python AI services with our responsive Node.js/React frontend required complex architectural problem-solving to ensure the "care loop" was seamless.

Accomplishments that we're proud of

Shifting the Dynamic: We built a tool that doesn't just "watch" old people; it actively reduces the cognitive load on caregivers.
Technical Empathy: We successfully implemented high-level AI (Clinical Decision Support concepts) not just for data, but to facilitate a more human, respectful interaction.
Peace of Mind: Knowing that this tool can turn a frantic "Where are you?" phone call into a calm "I see you're safe at the park" notification is a massive win for relationship quality.

What we learned

We learned that context is care. Mere data points like $ GPS, Timestamp $ are useless without the context of relationship dynamics. Technically, we expanded our toolkit with Computer Vision and OCR, but conceptually, we learned that the best technology is the kind that disappears, leaving room for human connection.

What's next for SafeStep

We plan to introduce "Mood & Social" metrics tracking not just if a senior took their meds, but if they’ve had social interaction today. We also aim to integrate with Electronic Medical Records (EMRs) to close the loop between the home, the caregiver, and the physician.