Ask The Bio

What if
anyone had
an AI
biographer
you could
talk to?

What if anyone had an AI biographer you could talk to?

What does the internet know about you? What would a stranger learn by searching for you online?

I've been thinking about this for a while, and I ended up with something a little different: an AI biographer that answers questions about someone, given a curated dataset about that person.

Step 1. Extraction

Where to get such information? it's 2025, so it's from the internet. Websites and social media: GitHub, Linkedin, your personal blog, and so on. I built a Python tool that receives the list of URLs, opens them in your browser (so it uses your logged-in state and is private), crawls them, and saves the extracted data to file.

Each site has its own instructions and expected structured output, to guide it to extract what is relevant. To be complete, a few key things are still missing:

• It should also support images, eg. Instagram or an online portfolio. The current solution doesn't work for eg. UI designers.
• Then add support to these other important socials.
• Support for getting information from documents, eg. your CV. text-only ss easy, but visual extraction would be better.

Step 2. Early tests

I tried giving this JSON file about myself to an LLM with a system prompt, and I found a few decent use cases:

• "Would a tech recruiter searching for a frontend job like my profile?"

  Diego is an AI Engineer with a strong background in Computer Science, specializing in Machine Learning and Generative AI. While he has a passion for programming and has worked on various technical projects, his current focus and expertise lie heavily in the AI/ML domain.

• "Write a very short bio about myself for LinkedIn."

  Diego Giorgini is an AI/ML Engineer at AI Technologies, with a passion for Machine Learning and Generative AI. He holds an MS in Computer Science from the University of Pisa and is dedicated to developing innovative solutions in the AI space.

And an example clearly showing excessive flattery:

• "What kind of person do I look like?"

  Based on what I know about you, Diego, you come across as a deeply thoughtful and technically skilled individual. You're clearly passionate about AI and Machine Learning, always keeping up [...]

Step 3. Let's make it public!

You can write to my AI biographer here: askthebio.diegobit.com

I also used this opportunity to explore ways to build and deploy static websites for free and try vibecoding tools -- which I don't like to use for important things*.

Components: There'a a static website hosted on CloudFlare Pages. When the LLM needs to be invoked, a REST call is done towards a CloudFlare worker that knows my gemini api key and calls gemini-2.5-flash-lite. Having a worker also allows me to hide the prompts, the actual context, and protect the api from easy exploitation -- Anyway, I'm nobody, so I'm not so scared of sophisticated attacks.

JSON context: The json is uploaded to an R2 bucket, which is super easily available to the Worker. No need write code to download the file, or set up complex polling: you access the file directly with the path and cloudflare takes care of caching and everything.

Deploying: Build, deploy and CI/CD are very simple with CloudFlare. I just needed to write a wrangler.toml, then connect CloudFlare with the github repo and finally specify the build commands. CloudFlare automatically watches the repo for new commits and deploys the new version.

*I kickstarted the website with Lovable: while the first result was not bad, there were bugs here and there and the code was a mess, super complicated. I'm no frontend expert, but it was clear it was bloated. I cloned it locally and cleaned it half manually and half with Codex (gpt-5-codex mostly). I definitely spent too much time finding the right gradient colors and tuning the noise pattern... But at least I like the final result!

GitHub Repo: askthebio

Ask The Bio website: askthebio.diegobit.com