Grafo | Devpost

💡 Inspiration

Grafo - γράφω (gráphō) means “I write” in Greek.

Writing cover letters can be an incredibly tedious task. It requires a lot of time and effort to craft the perfect letter for each job you apply for and research the company’s culture and mission statement. After a while, it can become quite tiresome to write up yet another version of your resume with slight variations in order to make it stand out from other applicants.

However, writing cover letters is essential to land your dream job or even getting noticed by potential employers. A good cover letter should show off your skillset in relation to the position you are applying for and demonstrate why hiring managers should take notice of you over all other candidates who have applied for that same role. If done correctly, this could be one step closer to getting hired! So I made Grafo

🖥What it does

Grafo is powered by a generative pre-trained model of cohere-ai to generate the cover letters needed. Users will write their prompts as detailed as possible and the web app will create the response in the best way possible.

⚙️How we built it

I used a technique called prompt engineering by using the cohere ai platform. Prompt engineering uses techniques from artificial intelligence (AI), natural language processing (NLP), and computer vision to create intelligent dialogue agents capable of generating appropriate responses in real-time without waiting for external input like voice commands or keyboard strokes. These agents are designed to mimic human conversation by recognizing patterns within conversations between people as well as understanding context clues such as grammar rules, slang terms, and cultural references so they can respond accordingly with relevant information while maintaining conversational flow throughout the interaction. For the front end of the website, I used React JS with Tailwindcss in styling. And for the backend, I used express js to get the API results.

🎢 Challenges we ran into

I wanted to make the output results to be in ideal letter in terms of the UI and to some extent that placed a challenge in the implementation process. The other one is hosting the express server for many users. As I used the trial keys for all endpoints, they were all rate-limited at 100 calls per minute.