I just finished building v3 of the chatbot on this website. The obvious thing to do would be to announce that it is smarter. I would rather ask you to try it.
Ask about my professional background. Ask how my engineering experience connects to enterprise sales. Ask about a role you are hiring for. Then ask a follow-up that makes the first answer work a little harder.
A useful front door
The bot has three jobs: answer questions about my professional life, help you request my resume, and suggest times for a conversation.
Those sound simple until you build them. A resume should follow the request and approval flow. Calendar suggestions need to come from actual availability. And an answer about my career should stay grounded in my experience instead of inventing an impressive-sounding version of me.
For scheduling, it can suggest three available 25-minute windows, account for your time zone, and respond to preferences such as Thursday afternoon. Google Calendar handles the final booking. A suggestion is not a reservation, so the booking page makes the final availability check.
The model is only part of the product
I used Amazon Nova for this project. In my own use case, it worked better for me than Gemini 1.5 Flash. That is a personal observation about this implementation, not a universal model ranking.
The work around the model matters just as much: which facts it gets, what tools it can call, how it handles failures, and where the system should stop guessing.
A friendly answer is useful. A friendly answer attached to the right action is much more useful.
Now I need real questions
Building something yourself gives you an unfortunate advantage as a tester: you know how it is supposed to work. Real visitors do not ask questions in the order I imagined.
So please try it. Give it a specific professional question, change your mind halfway through, or ask it to help you find a time. If something feels awkward or wrong, tell me.
I want this to be a useful way to start a human conversation. The best test is whether it helps you get to one.