Introduction
In the sea of hackathons, one of the most well-known in India is the Smart India Hackathon (SIH).
I and my team DelTechNeek took part in our college's (Delhi Technological University) internal hackathon. There were 180+ teams and 1200+ participants in the internal hackathon itself. We worked on a problem statement for the Ministry of Urban Affairs (SIH1516):
Suggest an AI-based solution to enable ease of grievance lodging and tracking for citizens across multiple departments.
TL;DR
We didn't make it to the shortlisted teams — but we learned a ton and ended up solving a real-world problem.
The Team
Sara Sanwal
B.Des (UX Design), Year III
Ekaksh Janweja
B.Tech (Civil Engineering), Year III
Aryan Sidhwani
B.Tech (Engineering Physics), Year III
Ayan Sajwan
B.Tech (Engineering Physics), Year III
Aakhyat Bagga
B.Tech (Engineering Physics), Year III
Abhinav Jha
B.Tech (Electrical Engineering), Year II
A Video Demo of the App
The Vision Behind GovBuzz
Introducing GovBuzz, an AI-powered chatbot for lodging grievances. It offers a simple voice command feature for language-inclusive access. GovBuzz processes complaints, filters spam, and directs them to the correct department. You get real-time updates and can participate in a user forum for transparent issue resolution.
Tech Stack
Mobile App
I was responsible for developing the mobile app, a critical component for user interaction. Our app required multilingual support and cross-platform functionality, leading me to choose Flutter for its versatility. For the database I opted for Firebase, driven by the need for speed during the hackathon.
Other technologies used in the app:
Riverpod
state management
Firebase
auth + Cloud Firestore
Routemaster
navigation
Dio
API requests to the Django backend
Django Backend
Aryan took up the task of making the backend to connect the ML and the Flutter app via APIs. Django was our first choice because the ML models were written in Python — integration became very easy. We also tried deploying the backend to AWS, but couldn't achieve this in such a short duration.
Deep Learning Models
Ayan and Aakhyat were responsible for creating the ML models. They collected the data from Google Forms and social media platforms such as Twitter and Facebook.
This data was preprocessed and then two models were created from it:
Spam Detection Model
detects whether a user input is valid for grievance lodging
Department Detection Model
classifies the user input into the most likely department
These models used Tensorflow, NLTK, and NLP techniques to process the user input.
User Research and Design
Sara, with a background in UX, did all the primary and secondary research and helped us identify the use cases. She made the designs for the app and the admin dashboard, and took charge of the presentation.
User Cases
GovBuzz catered to diverse user profiles, each with unique needs:
Urban Resident
report issues like potholes, malfunctioning streetlights, and garbage accumulation
Rural Farmer
seek assistance for irrigation, pest control, and crop advisory services
Small Business Owner
advocate for road connectivity and uninterrupted power supply
Senior Citizen
request better healthcare facilities, community centers, and safety measures
Admin Dashboard
Abhinav created a dummy website with a frontend. He used React and Tailwind CSS to make the admin dashboard fetch the grievances from the backend and route them to the concerned official/authority.
Business Model
The team imagined a subscription-based model where each ministry or department could subscribe to our service. GovBuzz would then tailor AI models to their specific grievance data, facilitating quicker problem resolution.
Why didn't we succeed?
Despite creating a full-fledged prototype and a complete problem statement analysis, these could have been some reasons we didn't make it through:
We will be back next year!