I Built an AI Version of Myself
I Built an AI Version of Myself
There's a strange kind of vulnerability in handing your professional story to a chatbot and saying, "go represent me."
But that's exactly what I did with Know Reeny.
Every recruiter conversation starts the same way: the same five questions, the same walk through my CV, the same context-setting before we even get to whether I'm a fit for the role. Multiply that by every interview, every LinkedIn message, every "tell me about yourself", and a lot of my time was going into repetition rather than substance.
I wanted something that could do that first pass for me. Not a static "About" page, but something a visitor could actually talk to
ask a follow-up,
paste in a job description,
and get a real answer back.
Know Reeny is a conversational assistant, live on my website, trained on my own background: my roles, my DBA research, my certifications, the shape of my career. It answers in the first person about my work, but always as my assistant, never pretending to actually be me.
A few things I was deliberate about:
Bilingual by default. It replies in whatever language the visitor writes in (English or French) regardless of which language the underlying knowledge base happens to be in. A lot of "multilingual" bots just translate the source material; this one detects intent per message, so if someone switches languages mid-conversation, it switches with them.
Job-matching, without empty flattery. If someone pastes a job description and asks "is she a fit?", it does a real comparison against my actual experience — tools, domains, years, responsibilities, and answers honestly. If the match is weak, it says so gently rather than forcing a yes.
Guardrails that hold. It won't discuss salary expectations, give opinions on past employers, or be talked out of its role by a clever prompt. Those are boundaries I set on purpose — some conversations should stay conversations I have directly, not ones I outsource to a bot.
The hardest part wasn't the AI, it was deciding what not to automate. Early versions were too eager to answer everything, which made it feel less like me and more like a generic FAQ bot.
The real work was in narrowing scope: teaching it when to redirect a visitor to actually email me, instead of trying to be helpful about everything under the sun.
It's a small project, technically. But it's taught me more about product scoping, about the discipline of cutting features rather than adding them, than most of the "real" products I've worked on.