MarkBot | Devpost

Halfway trough the project, high hopes
First sequenz diagramm
First pitch at 10:45am

Inspiration

Our goal was to solve more than one problem.
A polysoution for a polycrisis Marketing for non profits with specific goals require knowhow and experience from both domains.

Tracks

Social impact, Beginner, UI/UX

What it does

You enter your company data, and our locally hosted LLM generates a full marketing plan. You can then immediately refine it through an interactive conversation. Once the plan meets your needs, one click produces a polished, executable campaign PDF—ready to deploy.

How we built it

The frontend was built with React and Typescript and styled with Material UI. The backend was built with Python FastAPI and powered by Ollama's Gemma3:1b Large Language Model.

Challenges we ran into

LLM not saving Chat history. merge issues PDF creation LLM not fast enough Model expects most recent message from user and not from agent

Accomplishments that we're proud of

IT WORKS PDF creation

What we learned

Start earlier. Communicate what past projects teammates have done (Intro)

What's next for MarkBot

Implement a web scraper to gather even more information for clients Improve on personalized templates (individual template creation based on field and goal trough LLM) Fine tune model selection (maybe SLM)