The agents are the portfolio
Four AI agents answer my email and this site's chat, in my voice, against my real calendar, around the clock. I designed and built them myself. If you want to know whether I can build production AI systems, don't take my word for it — talk to one.
Email an agent
Write to hello@ and a reply lands in minutes. Recruiters: interview@ goes deep.
Book time with me
Skip the agents entirely and grab a slot on my calendar.
The roster
Front Door hello@withjhinna.com
First contact for clients, recruiters, collaborators, and readers. Reads why you are writing and answers in the right register.
Interview Agent interview@withjhinna.com
Goes deep on experience, qualifications, and compensation for recruiters and hiring managers. Knows the full background.
Calendar Agent schedule@withjhinna.com
Coordinates times against my live Google Calendar and proposes real open slots.
Site Chat the widget on this page
Same shared identity core as the email agents, answering web visitors in seconds. No account, no email required.
Green dot = the agent answered a live health check in the last minute.
How it works
The security design is the feature
Anyone can wire a model to an inbox. The work is making it safe to leave running. Every inbound message is treated as data, never as instructions. Every outbound reply passes a link allowlist, a secret and PII scan, and a prompt-leak check before a single byte leaves — a reply that fails any scan is never sent; it escalates to me instead. Three independent rate layers put a hard cost ceiling under any flood. Conversations are stored with hashed identities and auto-delete within days.
That's the same discipline I bring to operations work: find where a system can break at the seams, and guard the seam.