Talk2Doc

Inspiration

Definitely from my own experience, whenever I create a resume, I tend to compare with other people to see if my resume stands out or its better than theirs. Not only that, but also I sometimes don't know what to add as my content, so I try to find templates, but if I find templates I have to either pay or re-do the entire resume just to ensure it is correct. So I asked myself what would be the best way to re-create a resume based on the best that the internet has told, and make sure its editable the way I want it to be. Thats when I thought of Talk2Doc.

What it does

Talk2Doc (Phase 1 ) is an Al-driven resume creator. The user can choose to either choose a resume template ready in the site itself, or upload an Image of the resume and the AI will analyze the image, producing an editable resume. It is a FULL voice editor where you can talk to the AI to edit the resume (able to also manually input each section if the AI gets the wordings wrong).

How I built it

This was built using the Antigravity IDE. The entire application was made with ReactJS. Amazon Bedrock is used to connect client using the Access Key ID, Secret Access Key, Session Token, and Region. Not only that, Supabase is used to store (storage bucket) the uploaded images (templates) to ensure it can be processed.

Challenges I ran into

1. AWS integration: This is my first time integrating AWS into my software, therefor I had many issues because i did not know how to connect it with the Bedrock Client.

2. Supabase Storage Bucket: I have used Supabase before, but only the Database and RLS policies, this is my first time using the storage bucket.

3. AI Image Analysis: The current version still has some bugs because the image is not extracted completely and accurately. The current code does extract most of the information, but not exactly all of the information.

Accomplishments that I am proud of

1. Using Bedrock and Nova: This is my first time using Bedrock and I honestly think it is so cool and easy to integrate. I will definitely continue to use this for future projects.
2. Image extraction: To extract the image as much as possible into variables and converting to a JSON format took some time. There will still be development on this project to be able to extract accurately.

What's next for Talk2Doc

There is alot more feature I plan to add. For example, not only Resumes' but also documents such as PDF and others. Not only that, image extraction will also include fonts and colors.

Built With

• amazon-nova
• amazon-web-services
• antigravity
• bedrock
• react
• supabase