Sales engineer & solutions architect · Castro Valley, California
I've spent a little over twenty years on the technical side of selling software. Most of that time has been in IT service management, sitting between a customer's IT team and the product and working out whether the two actually fit.
I started in 2002 installing access control systems in government buildings. Since then I've written simulation software in C#, installed and configured ITSM platforms at customer sites, and worked as a sales engineer for TechExcel, Ivanti, and EasyVista, often in evaluations against ServiceNow. Lately I've been building MCP servers that connect Claude to tools like ServiceNow and Zendesk.
Employees pasting company data into unapproved AI tools, attackers working at machine speed, and AI agents nobody signed off on are now problems an IT team has to watch for. Sentinel is a detection console I built, with Claude, for managed service providers who need to answer three questions for their customers: are staff pasting company data into AI tools, is someone attacking us at AI speed, and are unapproved AI agents running on our systems.
It runs 17 detection rules across those three goals against a Microsoft Entra tenant, proxy and egress activity, and endpoint inventory. The rules do the detecting. An AI agent, optional and cloud-based, only explains the cases in plain language. When a case fires, Sentinel opens an incident in ServiceNow with what was seen, which rules fired, and a recommended next step, so the technician works it in the ticketing tool they already use.
The demo below runs against a lab tenant: a burst of failed sign-ins followed by a success on the same account, uploads to an AI service, a tenant-wide admin consent given to an app, and a local model runtime on a laptop, all landing as ServiceNow incidents. It also states its limits up front. Sentinel has no coverage for personal phones or home networks.
Demo recording, about four minutes, run against a lab environment.
A client's CTO didn't want to open up SAML for syncing users into Ivanti. Instead of pushing, I proposed a scheduled CSV export over SFTP that loaded straight into the user directory, and demoed it working. It solved his concern, and meetings with that team were easier from then on.
Working deals across a thirteen-state Western territory, I kept running into requests for project portfolio management that the product couldn't cover. I wrote the gap up for product development, and it was built. That write-up is credited in the PPM chapter of Kevin J. Smith's The Practical Guide to World-Class IT Management.
Much of my EasyVista work was head-to-head evaluations with ServiceNow. We won several of them by running proofs of concept on the customer's real processes and data instead of on a standard demo script.
B.S. in Computer Science, Cal State East Bay (Hayward). Anthropic certifications: Claude 101, AI Fluency, and Model Context Protocol (Introduction and Advanced Topics). Ivanti Administrator.
I read a lot of history, mostly the nineteenth century. I've followed the Warriors and the 49ers my whole life. I was born in Concord and live in Castro Valley with my family.
If you're hiring for presales or solutions work, or just want to compare notes on ITSM or MCP, email me at darrylsarkisian@gmail.com or book a 30-minute call.