Renova | Devpost

Inspiration

Every fitness app out there assumes you already know what you're doing. They throw macros, splits, and calorie trackers at you like you've been lifting for years. But what about the person who just decided today that they want to change? The one who doesn't have a gym buddy, doesn't know where to start, and has no plan?

That's who we built Renova for. We noticed that physical fitness and mental health are deeply connected. exercise is one of the best things you can do for your mental state, but no app actually bridges those two worlds for someone starting from scratch. Most people hit one moment of motivation, download an app, get overwhelmed, and quit. We wanted to catch that moment and make sure they don't fall off.

What it does

Renova is a fitness + mental health app designed for people who are starting alone and need everything in one place.

Onboarding that asks why you're here, not just your weight and height. We use that context to shape everything.
Personalized workout and meal plans based on goals, experience level, and accessibility needs. AI-generated or preset options. We explain the why behind the food, not just count calories.
Mental health journaling with mood tracking, prompted reflections, and AI-powered supportive insights after each entry.
AI Coach via OpenRouter that knows your profile and journey. Motivates without being corny, suggests without diagnosing.
Progress engine : streaks, a visual journey timeline, progress photo "memories" that resurface over time, and milestone celebrations.
Buddy matching that pairs you with someone on a similar journey so you're not doing it alone.
Live form analysis (POC) : a Swift native camera module through Capacitor that uses CoreML pose estimation to give real-time form feedback on exercises.

How we built it

Next.js 14 (App Router) for the full-stack framework, wrapped in Ionic + Capacitor for native mobile. ShadCN/ui + Tailwind for the frontend. AWS DynamoDB for the database, Cognito for auth, OpenRouter for AI features.

The live camera POC uses a custom Capacitor bridge plugin into Swift. AVFoundation captures frames, Apple's Vision/CoreML handles pose estimation on-device, and we compare joint angles against correct form ranges for basic exercises.

Team-wise, we split into four domains: physical health, mental health, progress/retention, and lead/integration. Everyone had their own branch and folder. Nobody touched anyone else's code. The lead handled all merges. We did research first, structured data second, built components third, integrated last. That workflow is the only reason four people shipped a full app in 24 hours without killing each other.

Challenges we ran into

Scope was the biggest enemy. We had way more ideas than hours. Ranking systems, social feeds, restaurant menu scanners we had to cut features we genuinely liked to protect the ones that mattered.

Getting the AI coach tone right took more iteration than expected. Too clinical and it sounds like a legal disclaimer. Too generic and it sounds like a fortune cookie. Finding the sweet spot where it actually feels like talking to someone who cares was harder than the technical implementation.

The camera pipeline had its moments. Bridging React to Swift through Capacitor, getting real-time frame processing working, and making CoreML inference feel responsive all took more debugging than we planned for.

Also — building mental health features responsibly. We're not therapists. Figuring out what an app should offer vs. what it shouldn't touch required us to actually think carefully instead of just shipping whatever sounded good.

Accomplishments that we're proud of

A user can sign up, complete onboarding, get a personalized plan, log workouts, journal their mental state, get AI reflections, track streaks, see their timeline, and get matched with a buddy. That's a full product loop in 24 hours.

The camera POC works. React to Capacitor bridge to Swift to CoreML to real-time feedback. For a hackathon, we'll take it.

Two of our teammates had barely built anything outside of class before this. By the end they were shipping real components. The collaboration model we set up made that possible.

What we learned

Researching before coding feels slow but saves you from building the wrong thing at 3am. Our "no code for the first two hours" rule is the reason we didn't have to throw anything away later.

How you organize a team matters as much as how you write code. Branch strategy, folder ownership, and component interfaces meant zero merge conflicts that actually cost us time.

AI is powerful but it needs a human shaping it. Every AI output from the coach to the workout plans was validated by someone who understood the domain. The AI multiplied our work, it didn't replace it.

What's next for Renova

Expanding the camera form analysis with a bigger dataset covering more exercises and body types. Deeper buddy matching with journey-stage awareness and timezone scheduling. Partnering with certified trainers and mental health professionals to validate all content. Wearable integration with Apple Health and Fitbit. And small moderated group challenges not leaderboards, because we're intentionally anti-comparison.