💡 Inspiration Every semester, college students face the same recurring nightmare:
Scattered schedules: Timetables and classroom locations vary daily across theory halls, lab complexes, and lecture auditoriums. Deadline panic: Assignments and project submission windows slip by when tracked across disjointed emails, LMS portals, and scattered sticky notes. High-friction tracking: When packing a bag or rushing out of a dorm room with five minutes until lecture, unlocking a phone, opening an app, and digging through menus is slow and inconvenient. We asked ourselves: What if managing your entire academic life was as frictionless as talking to a roommate?
"Alexa, what is my next class?" "Alexa, tell Campus Companion I have to submit DBMS assignment within 5 days." "Alexa, DBMS assignment is completed."
This inspired us to build Campus Companion — an ambient voice-first student assistant powered by Amazon Alexa and backed by a zero-key, 100% offline local academic AI mentor and an interactive web dashboard.
🚀 What It Does Campus Companion bridges voice-first ambient computing with relational academic tracking and intelligent study mentoring:
🎙️ Natural Voice Control via Amazon Alexa:
Timetable Lookups: Instant answers to "What is my next class?" including room numbers, building names, and real-time countdowns (e.g., "starts in 45 minutes"). Deadline Registration: Natural language assignment logging with relative day countdowns (e.g., "submit DBMS assignment within 5 days"). Automatic Daily Reminders: Schedules recurring notifications each morning until the assignment deadline or until marked done. Completion & Dismissal: Voice commands like "DBMS assignment is completed" immediately mark the task done and dismiss pending reminders. 🧠 Built-In Local Academic AI Mentor (Zero API Keys):
Provides concept explanations and exam tips for core computer science topics (e.g., Database Normalization, BCNF, Operating Systems Paging, TLB, and Deadlocks). Formulates personalized study schedules and prioritization suggestions based on current workload. Runs 100% locally and offline without requiring external paid API keys or exposing sensitive student data. 👥 Multi-Tenant Student Isolation:
Automatically maps Alexa device tokens (session.user.userId) to distinct student database profiles (e.g., Ravi vs. Priya). Guarantees strict privacy: students only access their own timetables, pending tasks, and personalized reminders. 🌐 Interactive Web Dashboard & Voice Simulator:
Built-in Web Speech API audio synthesizer and voice recognizer for browser-based testing without physical hardware. Live Alexa Skills Kit (ASK) JSON Inspector showing raw incoming requests and outgoing speech payloads in real-time. Visual weekly timetable grid, active reminder countdown chips, and task management cards. Temporal Time Travel Simulator: Fast-forward time by 24 hours or jump to custom dates to test recurring reminders and assignment urgency progression. 🛠️ How We Built It We designed Campus Companion with a modular, serverless-ready architecture:
┌─────────────────────────────────┐ ┌─────────────────────────────────┐ │ Amazon Alexa Device / │ │ Interactive Web Dashboard │ │ ASK Developer Console │ │ (HTML5 / CSS / Web Speech) │ └────────────────┬────────────────┘ └────────────────┬────────────────┘ │ (HTTPS POST /api/alexa) │ (REST API) ▼ ▼ ┌───────────────────────────────────────────────────────────────────────────┐ │ Flask Backend Engine │ │ ┌─────────────────────────┐ ┌────────────────────────────────────────┐ │ │ │ Alexa Handler & ASK │ │ Local Academic AI Engine │ │ │ │ Speech Synthesizer │ │ (NLP Parsing & Academic Concept Mappings)│ │ │ └─────────────┬───────────┘ └───────────────────┬────────────────────┘ │ │ │ │ │ │ ┌─────────────┴───────────┐ ┌───────────────────┴────────────────────┐ │ │ │ Schedule & Timetable │ │ Assignment Deadline & Reminder Queue │ │ │ │ Temporal Engine │ │ Dynamic Priority Calculator │ │ │ └─────────────┬───────────┘ └───────────────────┬────────────────────┘ │ └────────────────┼──────────────────────────────────┼───────────────────────┘ ▼ ▼ ┌───────────────────────────────────────────────────────────────────────────┐ │ SQLite Multi-Tenant Database │ │ (students, courses, class_sessions, tasks, reminders, sim_states) │ └───────────────────────────────────────────────────────────────────────────┘
- Tech Stack Voice User Interface (VUI): Amazon Alexa Skills Kit (ASK), custom interaction model (interaction_model.json), custom slot types (AMAZON.SearchQuery, AMAZON.DayOfWeek, CourseSubject), SSML speech formatting. Backend Framework: Python (Flask, Gunicorn, WSGI serverless adapter for Vercel). Database & Persistence: SQLite3 with multi-tenant relational schema, foreign key constraints, and cascading deletions. Frontend Dashboard: Semantic HTML5, Vanilla Glassmorphism CSS, and asynchronous JavaScript with Web Speech synthesis and recognition. Deployment: Pre-configured vercel.json with serverless Python runtime (api/index.py) and HTTPS webhook endpoint for the Alexa Developer Console.
- Mathematical Models & Academic Logic A. Dynamic Task Urgency & Priority Function To recommend which assignment a student should tackle first, the academic scheduler calculates a dynamic urgency score $\mathcal{U}(i, t)$ for task $i$ at current time $t$:
$$\mathcal{U}(i, t) = \frac{\omega_i}{\max\left(1, , \lfloor d_i - t \rfloor\right)} \times \left(1 + \alpha \cdot \mathbf{1}_{{d_i - t \le 2}}\right)$$
where:
$\omega_i \in [1, 5]$ is the assignment credit weight. $d_i$ is the deadline timestamp and $t$ is the student's current (or simulated) timestamp. $\mathbf{1}_{{d_i - t \le 2}}$ is an indicator function that triggers an emergency urgency boost factor $\alpha = 0.5$ when 48 hours or fewer remain. B. Built-In Academic Mentorship & Memory Calculations When students query the assistant about Operating Systems memory architectures, the academic engine explains and computes Effective Memory Access Time (EMAT):
$$\text{EMAT} = h \cdot \left(t_{\text{TLB}} + t_{\text{mem}}\right) + (1 - h) \cdot \left(t_{\text{TLB}} + 2 \cdot t_{\text{mem}}\right)$$
where $h$ is the Translation Lookaside Buffer (TLB) hit ratio, $t_{\text{TLB}}$ is the cache lookup time, and $t_{\text{mem}}$ is the physical memory access latency.
For Database Management Systems (DBMS), the AI validates functional dependencies for Boyce-Codd Normal Form (BCNF):
$$\forall (X \rightarrow Y) \in \mathcal{F}^+, \quad X \text{ must be a superkey of relation } R$$
🥊 Challenges We Faced Alexa ASK Natural Language Ambiguity & Slot Resolution:
Challenge: Alexa user utterances vary widely—from "I have to submit DBMS assignment in 5 days" to "DBMS due in five days" or "within a week". Solution: Built a flexible entity resolution layer in assignments.py that normalizes text numerals ("five" $\rightarrow 5$), relative time anchors ("tomorrow" $\rightarrow 1$, "next week" $\rightarrow 7$), and course acronyms ("Operating Systems" $\leftrightarrow$ "OS" $\leftrightarrow$ "OS Lab"). Temporal Testing of Multi-Day Reminders:
Challenge: Testing whether daily reminders trigger accurately over a 5-day lifespan would normally take 120 hours of real-world waiting. Solution: Implemented an isolated simulation clock (clock.py). Through the web UI, developers can click "Advance +1 Day" to simulate time progression, watch the deadline count down, and verify that reminders update dynamically. Multi-User Privacy on Shared Voice Hardware:
Challenge: In dorms or shared apartments, multiple roommates might share an Echo device. Solution: Developed a multi-tenant authentication mapping using the Alexa userId header (amzn1.ask.account.*). If Ravi is speaking, only Ravi's schedule is queried; if Priya speaks on her account, her separate schedule is resolved. Zero-API-Key Reliability & Offline Privacy:
Challenge: Cloud LLM APIs often introduce latency spikes, cold starts, rate limits, and billing hurdles, while sending private student schedules to third parties. Solution: Designed a deterministic, local academic AI mentor that parses intent locally and generates structured educational advice instantly without external API keys or cloud dependencies. 🏆 Accomplishments That We're Proud Of True Hands-Free Academic Flow: Adding an assignment, checking classroom locations, and completing tasks can all be done verbally in under five seconds while running across campus. Flawless Alexa Developer Console Compliance: Implemented strict Alexa Skills Kit JSON specifications with session management, reprompts, and simple card visuals for the Alexa smartphone app. Zero Cost & Infinite Scalability: The entire system runs without paid API dependencies and deploys smoothly on Vercel's serverless tier. Dual Experience (Voice + Visual): A cohesive experience connecting auditory voice responses on smart speakers with a live, responsive glassmorphism web dashboard. 📚 What We Learned Voice User Interface (VUI) Design: Spoken speech requires concise, high-density clarity. Unlike visual interfaces with unlimited space, voice responses must provide the most critical detail first (e.g., "Your next class is DBMS in room B-204 in 15 minutes"). Temporal State Management: Managing cascading relational schedules and countdown-driven reminder states across multi-day lifecycles. Serverless Edge Adaptation: Handling SQLite persistence and dynamic path resolution within ephemeral execution environments. 🔮 What's Next for Campus Companion Direct Canvas / Blackboard LMS Synchronization: Bi-directional webhook sync to automatically import new course syllabi and assignment drop-boxes. Proactive Alexa Notifications: Integrating the Alexa Proactive Events API to push spoken chime alerts 15 minutes before class departure time. Campus GPS & Walking Time Integration: Estimating transit times between campus halls to advise: "You should leave now; Hall C is an 8-minute walk from the dorm." Peer Study Groups: Voice-coordinated study sprints where classmates in the same course section can synchronize project milestones.
Log in or sign up for Devpost to join the conversation.