Autofiller AI

Inspiration

As students and working professionals, we spend many hours writing essays about our background and qualifications when applying for jobs, accelerators, clubs, and other opportunities. Especially as students at UC Berkeley, we found it is common for students to spend more than 15 hours in a single week simply writing essays about themselves.

What it does

Through Autofiller AI, you input your resume, LinkedIn, Twitter, and other relevant information you would like to share. Afterward, when you browse an online application form, such as the Founders Inc application, the tool autofills personalized responses to each essay question based on the information you provided earlier, saving you hours of time.

How we built it

We created an embedding model built on top of the OpenAI API, allowing us to generate text (essay responses) with a large amount of context (information inputted by the user). We used Retool for our frontend dashboard, where users input and update the information that feeds into the model, and Supabase for our database storing embedding vectors for each user. We also developed a browser extension with HTML, CSS, and JavaScript that receives our model's output from a Flask backend and autofills the essay questions on the page.

Challenges we ran into

Our most significant challenge was managing a web stack with several components. When developing the browser extension, it was initially challenging to autofill content on the page.

Accomplishments that we're proud of

Our biggest accomplishment is producing an embedding model that outputs personalized essay content that would be acceptable for the vast majority of job, club, and accelerator applications. We are also proud of building a platform that seamlessly integrates multiple APIs and cloud services together.

What we learned

Our biggest learning was to plan and visualize our tech stack early on, rather than adapting it as the project progresses. This would have saved significant time and confusion during our development process. We also learned how to use multiple new technology tools, including Retool, Supabase, and embedded generative AI models.

What's next for Autofiller AI

We will conduct user research to validate the utility of this platform and ascertain whom it could benefit the most. After prototyping an MVP, we will launch it to real paying users. In the long-term, we aim to use this concept to launch a venture scalable startup.

Built With

• css
• flask
• html
• javascript
• openai
• retool
• supabase