𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗪𝗶𝘁𝗵 𝗚𝗼𝗼𝗴𝗹𝗲’𝘀 𝗔𝗜 𝗦𝘁𝗮𝗰𝗸: 𝗧𝗵𝗿𝗲𝗲 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝘀, 𝗢𝗻𝗲 𝗕𝗲𝗹𝗶𝗲𝗳

I’m Darshan Nayak, 18, currently in my second year of an Integrated B.Tech in Computer Science and Engineering (AI and Data Science) at MIT World Peace University, Pune. While most people my age are figuring out their timetable, I’ve spent the last two years figuring out how far an idea can go before it turns into something people can actually use.

That chase has taken me through multiple national hackathon wins and a spot in the top 106 out of 85,000+ registrations at Google’s Solution Challenge 2026, putting me in the top 1% nationwide. Along the way I built three products, all running on Google’s AI stack, each one solving a problem I couldn’t stop thinking about. Here’s what they do and how they work.

𝗥𝗲𝗠𝗼𝘁𝗶𝗼𝗻: 𝗥𝗲𝗵𝗮𝗯 𝗧𝗵𝗮𝘁 𝗙𝗶𝘁𝘀 𝗜𝗻 𝗬𝗼𝘂𝗿 𝗣𝗼𝗰𝗸𝗲𝘁

Physiotherapy has one glaring problem. The moment a patient walks out of the clinic, nobody is watching whether they’re doing their exercises correctly, or doing them at all. ReMotion fixes that using the one device almost everyone already owns, a smartphone.

It uses MediaPipe Pose to track a patient’s skeletal joints in real time and calculate the angle of every movement as they exercise. That data goes straight to Gemini, which understands the context of what’s happening and speaks back personalized guidance and corrections, in English, Hindi, Marathi, Konkani and many other languages due to Gemini’s multilingual support. So if a patient’s knee angle is off during a squat, they’re not left decoding a chart. Gemini tells them, out loud, in their own language, exactly what to fix in the moment.

Gemini also runs Flex AI, the recommendation engine behind the app. It uses retrieval augmented generation over WHO discharge protocols to build exercise routines grounded in real medical guidance instead of a generic library. And when a doctor uploads a patient’s discharge sheet, Gemini handles the OCR and pulls out the diagnosis automatically, so nobody has to manually retype clinical data that already exists on paper.

ReMotion exists because rehab shouldn’t need a clinic visit to be done right. It just needs a camera and a model that actually understands good form.

𝗘𝗺𝗲𝗿𝗴𝗲𝗻𝗰𝘆 𝗖𝗼𝗿𝗲: 𝗧𝗵𝗿𝗲𝗲 𝗦𝗲𝗰𝗼𝗻𝗱𝘀 𝘁𝗼 𝗦𝗮𝘃𝗲 𝗮 𝗟𝗶𝗳𝗲

Emergency Core started with a simple observation about hotels and hospitality venues. When something goes wrong, medical emergency, fire, security threat, the information that matters most is scattered across guests, staff, and whoever finally makes the call for help. Every second lost to that scramble adds directly to response time.

It starts with a 4-button IoT panic device. One press, medical, fire, security, or distress, sends a structured payload straight to the backend. No training needed, no app to open first. From there, Gemini takes over, reading the incoming signal and generating a full situation brief in under 3 seconds along with a recommended action checklist, then pushing it to the right staff through Firebase Cloud Messaging. If nobody acknowledges it within 90 seconds, Gemini doesn’t just resend the same alert. It generates a contextual escalation message with real reasoning behind it and sends it straight to a manager.

Every update reflects live across a coordination dashboard through Socket.io, so everyone watching the incident sees the exact same picture in real time. Sitting on top of it all, Vertex AI analyzes incident patterns over time to surface risk insights, so a venue isn’t just reacting to emergencies as they land, it’s learning where its actual risk points are.

Get Darshan Nayak’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

This is the project that landed us in the top 106 out of over 85,000 registrations at Google’s Solution Challenge 2026. Building something that has to work correctly under pressure, with zero room for a slow or confusing response, taught me more about designing real systems than anything else I’ve built.

𝗦𝘁𝗲𝗹𝗹𝗮: 𝗕𝗿𝗶𝗻𝗴𝗶𝗻𝗴 𝗛𝗼𝘀𝗽𝗶𝘁𝗮𝗹 𝗖𝗮𝗿𝗲 𝗛𝗼𝗺𝗲

Stella is different from the other two because it isn’t software living inside a phone or a dashboard. It’s a physical robot, and every single unit is calibrated to exactly one patient.

Before Stella ever meets that patient, the treating physician uploads the discharge summary through a doctor dashboard. Stella parses the document and pulls out the diagnosis, the recovery timeline, medications and dosages, and any additional clinical notes, then that data is pushed securely to the specific unit assigned to that patient. Once the transfer is complete, it’s wiped from the dashboard for privacy.

From there, Stella runs continuously. It tracks heart rate and SpO2 in real time, streaming both to a guardian’s phone app and the doctor’s dashboard as live graphs. It checks in on patient responsiveness throughout the day, and if the patient goes quiet too long, it alerts both the guardian and the doctor immediately. It combines vitals with facial expression analysis through its onboard camera to catch signs of distress or pain, and when it picks something up, it calmly guides the patient through measuring their vitals while flagging the guardian at the same time.

Physiotherapy is where Stella really earns the name nurse. It watches and analyzes posture and form across 24 different prescribed exercises, logging every session, what was done, when, and how well, straight to the doctor’s dashboard, so recovery gets tracked without the patient logging a single thing themselves.

For the voice side, Stella runs on Vertex AI to handle its speech, from the welcome script it delivers the first time it meets a patient to the calm, spoken guidance it gives during a moment of distress. That’s what makes the interaction feel less like a machine reading out alerts and more like something actually built to take care of someone.

Three different products, three very different problems, but the same belief runs underneath all of them. AI is at its best when it disappears into the background and quietly handles the thing a person shouldn’t have to think about, whether that’s correcting a squat, dispatching a hotel’s security team, or watching over someone recovering at home.

Would love to hear what you think, or connect if you’re building in this space too.

Portfolio: https://darshann.vercel.app

LinkedIn: https://www.linkedin.com/in/darshan-nayak-636393346/

GitHub: https://github.com/darshannayak21

Medium: https://medium.com/@ndarshan507

#Meetthebuilders #GenAIAcademyAPAC #Cohort3