Skip to content

Ruthvik-1411/GenAI-Projects

Repository files navigation

Gen AI Mini Projects

This is a repo to implement some Gen AI usecases that seemed interesting. List of projects/ideas I plan to implement:

  • ChatBots
    • Basic ChatBot -> simple-gemini-chatbot
    • RAG Based ChatBot (basic) -> simple-rag-chatbot-with-mesop-ui
    • RAG Based ChatBot (Complex datasets, using different techniques)
      • Diverse domains, something llm's are not trained on.
      • Various File formats, PDF, Docx, HTML, images, txt. Multimodal embeddings.
      • Rich enough data for retrieval + generations + evals using different techniques for chunking, retrieving...
      • Always on the lookout for such datasets.
  • NL2SQL
    • Basic NL2SQL Bot on simple data with viz.
    • Agentic N2SQL Bot
      • Generate insights, decompose queries, recommendations using adv techniques
      • Use code execution where required. Custom code exec can also be explored.
  • Webscraping
    • Webscraping/Research Agent to generate reports.
      • Kind of like deep research, ref: anthropic blog
      • Citations, report generation etc.
  • Media Generation
    • Youtube Thumbnail Generator -> yt_thumbnail_generation
    • Short story generator with images.
      • Generates image slides to depict a story. Ensure character consistency.
    • Short Video generator (max 30s). (Costly with veo, but good exploration)
  • Live Application (Multimodal Live)
    • Boilerplate code for multimodal applications for prototyping. -> Gemini Live BoilerPlate
    • A simple bank teller agent that helps gather some information about a/c balance, payments made etc (access to tools)
    • Live gent with access to some data and bunch of tools.(complex)
  • Agentic Applications (explore MCP along the way)
    • Website designing agent integrated with some IDE like copilot or via UI.
      • Creates and saves files, edits them based on requirement. Kind of like Replit.
    • API Integration assistant native to an IDE
      • Helps in providing code references using docs of python package or library or some API docs.
      • Modifies source code, create git branches etc to make edits in the code.
  • I'll add more ideas as they come up.

Note: These implementations are just to have a hands on experience with the latest models and frameworks. The code might not always be written as per best practices, but I try to do as much as I can. I will also be adding any learning for topics that I found interesting when implementing these mini projects.