Inspiration
The idea for LaunchPad AI was born directly out of my daily life as a parent. Every morning and evening, I found myself guiding my 8-year-old son through the exact same routines: checking his timetable, reminding him what to pack in his school bag, and sitting with him to keep him focused and motivated during study time. I realized that building these executive functioning skills—like organization and focus—takes immense patience and repetition. I wanted to create a tool that fosters independence in a fun, engaging way. Instead of a rigid to-do list app, I envisioned an always-patient, voice-native "coach." LaunchPad AI was built to give kids that extra layer of motivation and empathy, helping them take charge of their own learning and bag-packing routines, while giving parents a much-needed helping hand.
What it does
LaunchPad AI is a comprehensive student dashboard centered around a zero-latency, voice-native AI Coach. The Morning Check & Nightly Pack: The AI proactively reads your schedule. Ask "What do I need today?" and it tells you exactly what to pack based on your specific classes and homework notes. Dynamic Affective Dialogue: The Coach listens to how you speak. If you sound overwhelmed, it uses an Empathy Protocol to validate your feelings and break tasks down. If you need motivation, it enters "Hype Mode" to deliver a locker-room speech. Study Sprints & Gamification: Instead of just telling you to study, the Coach initiates "Study Sprints." Completing these sprints automatically awards you Focus Points and builds your Daily Streak, displayed in a sleek Gamification Widget to keep you coming back. Proactive Exam Alerts: The system constantly monitors your upcoming exams and will proactively interrupt to warn you if a test is exactly 3 days away, suggesting a custom review plan.
How I built it
Frontend: I built a highly responsive, animated UI using React, Vite, and Tailwind CSS. I utilized Framer Motion for fluid page transitions and stagger animations to give the app a premium, native feel. Voice Engine: The core of the app is powered by the Gemini Multimodal Live API (@google/genai). I established a WebSocket connection to stream raw PCM audio back and forth, enabling natural, interruptible conversations. Context Injection: I built a custom state management system that injects the student's live data (timetable, exams, current streak, and focus points) directly into the Gemini system prompt in real-time. This means the AI always has perfect context without needing to make slow external API calls.
Challenges I ran into
Prompt Engineering for Voice: Writing a system prompt for a voice agent is completely different than for a text chatbot. I had to heavily engineer the prompt to force the AI to use short, punchy sentences, pause frequently, and avoid reading long, robotic bulleted lists. Handling Interruptions (Barge-in): Ensuring the AI stops talking the millisecond the user interrupts was a challenge, requiring precise handling of the Live API's interruption signals and audio queue clearing. State Synchronization: Syncing the React state (like the Study Timer and Gamification points) with the AI's awareness required setting up custom event listeners so the AI always knows exactly how many points the user just earned.
Accomplishments that I am proud of
The "Elite Coach" Persona: I am incredibly proud of how distinct and motivating the AI's personality is. It truly feels like talking to a mix of Ted Lasso and a high-performance mentor. Zero-Latency Feel: By optimizing our audio buffering and leveraging Gemini's Live API, the conversation flows naturally without the awkward pauses typical of older voice bots. Beautiful UI/UX: The shift to the energetic Indigo/Violet color palette combined with Framer Motion makes the app feel like a top-tier, production-ready EdTech product.
What I learned
I learned that empathy is a technical feature. By explicitly programming an "Empathy & Validation Protocol" into the LLM's instructions, the user experience transformed from a transactional tool into a supportive companion. I also learned the intricacies of the Web Audio API and handling raw PCM data streams in the browser.
What's next for LaunchPad AI
LMS Integration: I plan to integrate OAuth for Google Classroom and Canvas so the app can automatically pull in assignments and due dates without manual entry. Parent/Teacher Dashboard: A companion view where mentors can see a student's Focus Points and Study Streaks to offer real-world rewards. Mobile App: Porting the React codebase to React Native so the Coach can live in the student's pocket for on-the-go motivation.
The Math of Focus
I realized that telling an 8-year-old to "go study for an hour" is overwhelming and abstract. To build real executive functioning skills, I needed to turn focus into a tangible, rewarding game. I built a custom gamification engine that calculates progress using what we call The Math of Focus: Micro-Sprints (The Base Rate): Instead of long, grueling study sessions, LaunchPad AI encourages short, highly focused "Study Sprints." Every minute of uninterrupted focus during a sprint earns the student Focus Points. It shifts the goal from "finishing homework" to simply "winning the next 10 minutes." The Consistency Multiplier (Streaks): Cramming doesn't work. To encourage daily habits, I implemented a Daily Streak system. If a student logs in and completes a sprint or a morning check every day, their streak increases. Executive Function Bonuses: Students don't just get points for studying; they get rewarded for organization. Completing the "Nightly Bag Pack" routine or doing a proactive exam review (prompted by the AI 3 days before a test) awards massive bonus points. The Result: By quantifying effort, I change the psychological framing. A student isn't just doing math homework; they are "leveling up their brain," protecting their 5-day streak, and earning points to show off to their parents. It turns the friction of daily routines into a highly engaging, rewarding loop.
Built With
- api
- canvasapi
- css
- dynamiclanguagedetection
- framermotion
- functioncalling
- geminimultimodalliveapi
- google-calendar
- html5
- lucidereact
- multilanguagesuport
- node.js
- react19
- server
- tailwind
- typescript
- v4
- vite
- webaudioapi
Log in or sign up for Devpost to join the conversation.