Inspiration

Our Inspiration for this project was our frequent presentations/pitch decks that we do in our college life and our work. we thought why not transform this experience into a dynamic experience that takes into account not only what the presenter wants to present but what the viewer wants to see as well

What it does

Our Agent at startup receives the knowledge base (documents) of the project/startup that will be presented and a brand guideline that defines the style of the presentation (primary, secondary, accent colors, diagrams style, fonts). Based on this the Agent generates initial problem statement slides to define the problem the product solves then it begins receiving questions from the viewers and generates stylized slides that answer these questions along with audio that makes the experience more interactive and adds context

How we built it

We built our system through Gemini flash and React components for Slide generation, Gemini tts for audio generation, a Rag pipeline for indexing and retrieving the knowledge base, services for orchestration of the whole process, FastAPI for endpoints and React for clean interface

Challenges we ran into

The main challenge we rain into is how to distinguish our project from services that provide slide generation while at the same time provide value ourselves. This made us bring our idea to live which is initial problem statement slide generation as a hook and then the rest of the slides are generated based on what the audience wants to see and hear, not just what the presenter wants to present

Accomplishments that we're proud of

We're proud of the product that our Agent came to be from just simple components that we built together over the past week and a half. Clean stylized slides generation, audio that adds interactivity and context. a seamless experience built from simple libraries and services

What we learned

We learned a lot about how to build systems that work instead of just streamlit demos. This was our first time using React so we learned a great deal about it. we learned how to restrict an LLM to certain output formats and output types. We learned how to orchestrate multiple components and services together to produce value

What's next for Pitch Deck Agent

Next we plan to make pitch deck agent production ready through: Efficiency and latency optimization Edge cases error handling Traceability and monitoring

Built With

Share this project:

Updates

Submission history