Work
Programs I have led inside banks and regulated companies, where the standard was sign-off from security, risk, and the people who run it afterward. And products I have built myself, where the standard is that they ship, stay up, and get used.
Five years leading a cumulative $11M portfolio of more than twenty cybersecurity projects at a Fortune 500 auto-finance company — and, alongside it, the security and governance track of the company's enterprise AI and machine-learning adoption program.
Each of these started as a problem I wanted solved and ended as software real people use. They are where I learned what it takes to get an AI idea into production — and to keep it there. Most were built with Claude as a development partner.
Quality management for regulated manufacturers: document control, CAPA, tablet inspections, supplier scorecards, training matrices, and on-demand audit-binder export. 21 CFR Part 11 ready.
What it shows: compliance-grade software on a typed data model designed for AI integration.
AI website builder: a complete site — layout, copy, images, colours, fonts — from a business description, with a visual editor, custom domains, e-commerce, and an AI blog writer.
What it shows: a full generative pipeline turned into a product people can edit and publish from.
Scheduling workspace for people and teams: booking links, Smart Overlay availability, round-robin and collective scheduling, routing forms, two-way Google and Outlook sync, SAML SSO and SCIM, webhooks, an Android app. It runs the booking link on this site.
What it shows: enterprise identity and calendar integration done properly.
AI voice receptionist: answers calls around the clock, books appointments, answers FAQs, routes and warm-transfers calls, qualifies leads, and follows up by SMS, WhatsApp, or email, in more than thirty languages. Integrates with Twilio, Salesforce, HubSpot, and others.
What it shows: real-time voice AI wired into the CRM and telephony a business already runs.
Donor-development system for nonprofits: donors classified by tier and capacity, action plans, ask-timing rules, and an AI Development Officer that drafts, never sends. Role-scoped access with audit trails and row-level-security isolation.
What it shows: human-in-the-loop AI on sensitive data, with the controls built in.
Local-first markdown knowledge workspace with an AI co-thinker: daily briefs, Story Mode drafting with citations, markdown kanban, a built-in MCP server, bring-your-own model, Obsidian-compatible.
What it shows: AI that works over your own files, locally, with citations.
AI-assisted operations platform for Islamic centers: communications, digital signage on Raspberry Pi, program registration, scheduling, volunteers, financial aid, and service tickets, from one command center.
What it shows: a multi-module operations platform with an AI work queue and approvals.
Financial-aid application portal for community organisations: a four-step, no-account application with reference-number tracking, caseworker review, and four languages.
What it shows: a sensitive-data intake system built for people who are not technical.
AI career platform: resume tailoring with verified claims, an AI mock interviewer with role-specific questions, and application pipeline tracking.
What it shows: structured AI coaching with claims checked against the source document.
Thirty minutes. Bring the problem; I will tell you what I would build and how it would get through your security review.