Inspiration
Hackathons are fast-paced and exciting, but many participants—especially beginners—struggle to decide what to build, whether their idea fits the rules, and how to execute it within limited time. We noticed that a lot of good ideas fail not because of lack of skill, but because of poor planning and uncertainty. BuildBuddy AI was inspired by the idea of creating a friendly AI co-founder that can guide anyone, even a beginner, from a hackathon idea to a deployed project with clarity and confidence.
What it does
BuildBuddy AI takes the name of a hackathon as input and understands its theme, rules, constraints, and judging criteria. Based on this, it recommends five project ideas that are feasible, high-impact, and quick to build. Once the user selects an idea, BuildBuddy AI generates a visual execution flowchart, breaks the project into simple step-by-step tasks, explains the required file structure and code in easy language, and finally guides the user through deploying the project.
How we built it
BuildBuddy AI is powered by the Gemini 3 API. Gemini 3 Pro is used for long-context reasoning to analyze hackathon details, rank suitable project ideas, and plan multi-step execution workflows. Structured JSON outputs are used to represent tasks, dependencies, and progress. Nano Banana Pro is integrated to convert these structured plans into visual flowcharts and architecture diagrams, making the agent’s reasoning transparent and easy to understand.
Challenges we ran into
One of the main challenges was simplifying complex technical concepts into explanations that feel intuitive and beginner-friendly without losing correctness. Another challenge was ensuring that the generated project ideas were both realistic to build within hackathon timelines and aligned with the rules and judging criteria.
Accomplishments that we're proud of
We are proud of building an autonomous system that goes beyond idea generation and actively guides users through planning, building, and deploying a project. The visual flowcharts generated using Nano Banana Pro make the AI’s reasoning visible, which helps users trust and understand each step of the process.
What we learned
Through this project, we learned how Gemini 3’s advanced reasoning and long-context capabilities enable true agentic workflows. We also learned the importance of explainability in AI systems and how combining reasoning models with multimodal visual generation significantly improves usability and clarity.
What's next for BuildBuddy AI
In the future, BuildBuddy AI could support team collaboration, multiple tech stacks, real-time progress tracking, and deeper deployment automation. We also plan to expand support for different hackathon platforms and provide personalized guidance based on user skill level.
Log in or sign up for Devpost to join the conversation.