-
-
AuditBot web app landing view with primary new analysis workflow (category, target role, file upload).
Inspiration
This idea came from my own struggle. I kept seeing strong applicants (friends, classmates, even myself) get rejected from opportunities without ever really knowing why. The feedback, if it came, was vague or extremely polite, and by the time you realized what went wrong, it was already too late to fix it. Over time, I noticed the patterns, which were vague narratives and small but fatal mistakes that kept repeating. I wanted something that would catch those issues early, a tool that feels like a strong first reviewer who tells you what others won’t. Audit Bot delivers direct, brutally honest feedback, giving a clear “likely accepted” or “likely rejected” verdict along with clear ways to improve.
What it does
Users begin by selecting an audit category, such as a resume, insurance claim, scholarship, college application, hackathon pitch deck, or video pitch. Then, they enter their target role along with the context and upload their file (resume, video, or any type of related files). The system uses Gemini, optimized with system instructions and output, to analyze the content. The output includes a verdict, key strengths, a reasoning summary, red flags, a three-step rescue plan, and a logic check tailored to the document type (like ATS feedback for resumes). Signed-in users can save their audits and revisit them later, powered by Firestore.
How we built it
We built the app using React, TypeScript, and Vite for speed. Gemini is connected with a fixed prompt and structured format to keep results consistent. Firebase handles login and data storage. The design uses a dark, bold style to match the no sugarcoating idea.
Challenges we ran into
It was tricky to keep the AI responses consistent with our desired response, handle video analysis smoothly, support different file types, and balance brutally honest feedback with helpful advice.
Accomplishments that we're proud of
We created a clear product identity and a smooth flow from upload to results to saved history. The app supports multiple document types, including videos, and keeps outputs consistent across the UI.
What we learned
We learned that good prompts and structure are key to reliable AI results. We also learned that the design of the program is the key to delivering to the user the best experience, as design determines the function.
What's next for Audit Bot
Next steps include adding a backend proxy for security, export/share options, before-and-after comparisons, AB testing, smarter history search engine, and an optional tone slider for controlling the intensity of the feedback.
Log in or sign up for Devpost to join the conversation.