Roasted Beans - How to turn LinkedIn profiles into personalized comedy gold

Roasted Beans is a research project by Kaffie that explores AI-powered automation for creating personalized content at scale. The concept? Take someone's LinkedIn profile and automatically turn it into a sharp, humorous roast video, completely hands-off from start to finish.

This wasn't about building the next big startup. It was about answering one core question: can AI create content that feels genuinely personal and humorous without any human touching the process?

The Project

This project aimed to test whether we could build a fully automated system that creates personalized content for each individual user. In other words: turning boring LinkedIn data into entertaining, custom-made roast videos that feel like they were written specifically for that person.

Let's make this tangible:

Most AI content feels generic and templated. So before we dive in, let me explain exactly what we built and how it works.

We start with data, in this case: a LinkedIn profile URL. From there, a chain reaction of automation kicks in:

• Someone buys a Roast Key and shares their LinkedIn profile
• Our system scrapes their profile data automatically
• AI analyzes their job history, posts, and professional persona
• A custom roast script gets written specifically for them
• An AI voice reads the script while an avatar presents it
• Everything gets combined into a personalized video

The entire process looks something like this: Raw LinkedIn Data → AI Analysis → Personalized Script → Automated Video Production → Delivered Roast

How this automation turned out for Roasted Beans

Based on the LinkedIn profile data, our system creates several components automatically:

The first component is data extraction. We pull in job titles, company history, posts, connections, everything publicly available on their profile.

The second component is AI content creation. OpenAI analyzes all this personal data and writes a custom roast that references their specific career moves, industry, and professional quirks.

The third component is automated video production. ElevenLabs converts the roast into speech, D-ID creates an animated presenter, and Puppeteer generates a LinkedIn profile scroll video.

With this system, we could deliver:

• Unique content for every user - no two roasts are the same
• Fast turnaround - most videos ready within 3 minutes
• Zero human intervention - from purchase to delivery, it's fully automated
• Scalable personalization - can handle multiple orders simultaneously

The outcome

The real value of Roasted Beans was proving that AI can create genuinely personalized content at scale. With 30+ roasts delivered, we learned that:

• AI can capture individual personality from public data alone 💡
• Full automation is possible for complex, creative content production 🔧
• Users are genuinely surprised by how personal the roasts feel 💌
• Scalable personalization works - each roast feels hand-crafted despite being fully automated 📈

The research showed that we can move beyond basic personalization (like inserting someone's name) to creating content that genuinely reflects their individual professional story and personality.

What we can do with this

The automation we built for Roasted Beans opens up bigger possibilities for businesses. Imagine applying this same approach to:

• Personalized client messages - Instead of generic newsletters, create individual stories for each client based on their data and interaction history
• Custom sales outreach - Generate unique pitches that reference a prospect's specific company situation, recent achievements, or industry challenges
• Data-driven storytelling - Just like we did with anyKrowd (turning raw transaction data into clear insights), we can now go from customer data to personalized narratives that resonate with each individual

The core principle remains the same: take raw data about someone and automatically transform it into personalized content that feels genuinely crafted for them. Whether it's a roast, a sales pitch, or a customer success story, the automation handles the personalization at scale.

Side note: Yes, this helped fund a Bali trip, but the real gold was learning how to build truly automated, personalized content systems.

‍