Introduction ✨
Let's be honest-crafting personalized cold emails is NO FUN. I mean, we all know we should customize each mail, but who’s got time for that? That's the motivation behind Forger AI.
Now, What is Forger AI?
Well, It's a tool I created to automate the art of cold emailing, making you sound like you put effort into it (even when you didn’t). In this blog, I’ll break down how I tried out multiple APIs, ran into more issues than I thought I would, and somehow ended up with a solution that works lol.
Checkout the demo here: Youtube Demo Video
The Motivation behind Forger AI 🔥
Why did I build this? Simple—I hate manually writing cold emails. With Forger AI, the idea was to automate email generation in a way that felt personalized and human. The goal? Paste a job post URL, and yayy!—AI whips up a well-crafted email that is relevant.
But, of course, I couldn't just stop at generating one email at a time. Why not send hundreds in bulk while I’m at it? This is where APIs like Llama3.1 and SendGrid API come into play.
The Tech Stack:
- Frontend: Built with Next.js and TypeScript for a clean, scalable interface.
- Tailwind CSS: For styling the interface
- Backend: FastAPI paired with Langchain to handle the logic and API routing.
- Llama API: The heavy lifter for parsing job descriptions and crafting custom email content.
- SendGrid API: Sends all those beautifully generated emails into the void (aka recruiter inboxes).
- Stripe API: Experimented with adding a payment gateway for premium features, and Stripe was the go-to solution for handling smooth payments.
- GroqCloud: Manages CSV uploads for bulk recruiter emails
Architecture Diagram
Play around with the code here on my Notebook: email_generator.ipynb
✨ Features:
Cold Email Generation
You can generate humanish-personalized cold emails from job post URLs.
Just enter a URL and AI will take care of the rest lol
Send Emails to Recipients
Send the generated emails directly to the recipient’s inbox. No need of
Copy-Pasting.
Upload Portfolio CSV
Upload a CSV file with your portfolio and attach it to your emails
automatically. Saves time!
Bulk Recruiter Upload
Upload a CSV file with recruiter emails, and we’ll send your cold
emails to all of them at once.
Premium Features
Unlock extra features like bulk email sending and advanced
customizations with a premium upgrade.
Edit Email (Coming Soon!)
Currently working on editing the AI-generated emails before sending.
Email Preview
See a live preview of the email generated by AI before sending the
mail.
Copy Email to Clipboard
Copy the generated email with one click and use it wherever you want.
THE API PHASE!🔥
This phase involved a lot of experimentation and broke a few things before finally selecting one.
I tried out OpenAI models ( GPT-3/GPT-4 ) and though it was super-versatile, it's pretty generic and the cost of using GPT-4 for large-scale tasks adds up pretty quickly. Then we went on with Claude and it was also quite similar.
API/Model | Pros | Cons | Why Not? / Why Chosen |
---|---|---|---|
GPT-3 / GPT-4 (OpenAI) | - It's obviously versatile, powerful & handles a wide range of tasks. - Well-established ecosystem and integrations. - Generates large, cohesive content. |
- But it's too formal and generic. - Too $$$ for large-scale use lol. - Overkill for specific cold emailing tasks. |
- Emails felt too “corporate” and lacked the human-ness?. - I was looking for something more focused for job-specific cold emails. |
LLaMA API | - Context-aware, produces personalized humanish emails. - Natural, conversational tone. - Requires less fine-tuning. - Fast and efficient performance. |
- Newer ecosystem, so lacks the broader integrations. - Slightly less community support compared to GPT though. |
- Finally chose this for efficient, human-like, job-specific email generation without the need for excessive fine-tuning. |
Claude (Anthropic) | - Ethical AI design, prioritizes safety. - Great for general-purpose tasks, such as conversation and customer service. - Fast response times. |
- Lacked the customization needed for cold emails. - Less flexibility in fine-tuning for specific, tailored content. - More focused on broader use cases. |
- Didn’t meet the personalized email requirement and wasn’t as adaptable to job-specific cold emails. |
GroqCloud API | - High-performance AI infrastructure. - Optimized for large-scale tasks and ML model deployment. - Efficient use of resources for certain operations. |
- Overkill for this project’s focus. - More suited for ML tasks rather than personalized email generation. - Complex setup. |
- Too complex for the goal of generating cold emails. - More focused on infrastructure and ML tasks, which wasn’t necessary for this project. |
Future plans for Forger.AI:
For now, I'm working on adding AI Agents that can handle email-follow ups and responses which AGAIN, can aid my laziness lol. Also, right now, Forger AI personalizes emails based on job postings. Next up, I'll be integrating LinkedIn and other sources to craft emails tailored to individual profiles. Currently improving the premium feature ie, the bulk recruiters email data for automating email sending process.
Final Thoughts <3
Building Forger AI was a journey of many APIs, lots of coffee, and a few facepalm moments. Well atlast, we made it though, a solution to automate the boring cold-email content process. There's still more to come though, more features and more learning!
If you read till here, thank you so much and I hope it was interesting!
Contact me 😎
Email: juanitacathy486@gmail.com
Github: JuanitaCathy
Top comments (0)