Watchdog

Inspiration

As students, we actively seek job opportunities to help grow our skillset and embrace learning opportunities. Unfortunately, the job market is also a place where scam artists try to exploit young professionals and steal their personal information resulting in devastating consequences like identity theft. This issue does not only affect us, but many other job seekers in the market. Even beyond the job market, we see misinformation on social media platforms, emails, and SMS. That is why we created Watchdog – to help combat scams and information in the age of rapidly growing technology.

What it does

Watchdog is both a web application and a chrome extension that analyzes text to detect any malicious intent from the sender, whether it be scams, misinformation, AI-generated, or a combination of these. Watchdog makes it easy to detect scams with just a click of a button with our browser extension – no hassle!

How we built it

Watchdog was built with Svelte, a web framework that makes it easy to create web applications using only vanilla HTML/CSS and JavaScript. Svelte was using both our web application and chrome extension, and it uses a cutting-edge large language model to perform evaluations to help you surf the internet with peace of mind.

Challenges we ran into

The task of prompt engineering was the primary challenge that our team faced. To get desirable responses, a large amount of finetuning the prompts was required. In addition, we employed prompt engineering techniques like setting context, allowing the model to think by providing examples of scams and misinformation, and crafting the prompt with specific syntax to allow the model to discern types of information. After a few hours of tuning, we were getting the responses we wanted!

Accomplishments that we're proud of

Our biggest accomplishment was completing all of our objectives that we initially set out to complete by providing a complete experience for the user with both a cross-platform web application and a chrome extension. This was a large team effort, and we are very excited to share the results of our collaboration with everyone!

What we learned

Everyone on the team picked up new technical skills, which will help each of us expand our toolset for future applications and development.

What's next for Watchdog

There are a few features that we still want to implement. One of them being able to grab text from an image and analyze the image rather than requiring the user to supply a text input. Finally, we also want to be able to analyze entire webpages and news articles to provide an even more practical experience of our application.

Built With

• chrome
• css
• figma
• gpt
• html
• javascript
• openai
• sveltekit
• vercel