Project Saadhna — Building AI Course Recommender on GCP
The right educational resource? Get recommended!
Overview
Students often fail to choose the right educational resource like course, internship, events, projects, including Project Saadhna and Code Vipassna. Students sometimes are influenced by friends, spam emails, advertisements that aren’t the best in the market, sometimes students even end up paying thousands of rupees for internships!
The AI — Chat bot is trained on a custom data-set prepared by my team, to recommend actually industry standard, useful educational resources, courses, events.
We consider high quality industry standard courses like Google Cyber Security Professional, The Odin Project for web development, cs50x by Harvard for Computer Science basics.
AIM
- To guide students to events like code vipassana, workshop meetups held by clubs like Azure developer club, expert good quality courses like cs50x, Google Cyber Security Professional Certificate, etc.
- To stop students from choosing internships where they pay themselves, choose training programmes that use recorded videos charging like 30k per student.
Design
The AI Chat — Bot is hosted on the Google cloud platform, built using Vertex AI Agent Builder. The chat bot runs on a custom data set, stored on a google cloud bucket.
Prerequisites
- A web browser to access the AI-Chat Bot’s website.
- A internet connection.
- Access chatbot
Building Process
This project initially started as a Project Centric Learning’s project, which me and a team of 5 started working on in my university. We discussed this idea with our project guide and explained the need and explained the cause on how our peers choose the wrong course and ended up paying over 30k for an internship which wasn't quite useful.
- Dataset creation
To start the process, we listed the flow-through we would be following for the project. Initially we needed a quality dataset which we will be using for the recommendation, built through experiences, industry standard, expert recommendations, etc. While initially we have our dataset as a pdf of instructions and course details, we plan on expanding it towards a Big Query table.
2. Google Cloud Data Store
To further use the dataset, we uploaded the dataset into a google cloud bucket and created a data store.
3. Vertex AI Agent Builder
Using Vertex AI Agent builder, we created the chatbot. Used the datastore to train the chatbot and set the grounding as very low for accuracy.
4. Deployed the agent to my website using Dialogflow Messenger
Dialogflow Messenger created the required API Script that I used on the website which will be hosting the chatbot to recommend resources and be available to the end users.
Result
The chatbot can be accessed and tested on https://nihalhu.co/Course-Recommender. Start with a hello or directly ask for courses on topics like digital forensics, Finance, Cyber Security, Tally, blockchain, etc.
Try: “Recommend me a course on Cyber Security”
Note: The dataset is still in improvement and will take some time for additions and optimization.
What’s Next?
As of today, the chatbot recommends a few courses selected by us. Future we hope to expand this to internships, events like code vipassana, hackathons, local university events, more courses, projects and help users learn open-source contributions.
Choose the right educational resource, do not fall for fake internships and utilize your time and effort for the right resource!
Links
University Team — Details
- Nihal H U
- P Om Shiva Prasad
- Tharun Agarwal
- Vanshika Shri N
- Varshini Krishnamurthy
University Guide: Assistant Professor — Sushma B S, Jain University
Project Saadhna Guide: Sarvesh Shashikumar
