Work
Things I've built that are still running.
Client systems first, then the AI tooling I run my own practice on. Clients are anonymized. Everything here is real work.
Client case studies
Usage-based SaaS
Phased contracts the billing tool couldn't do
Sales was closing two-phase deals that the billing platform had no concept of. Designed a linked-contract workaround, built production templates, and connected Claude so it drafts both contracts from the SOW for a human to approve.
B2B healthtech
Every inbound lead got the wrong email
Anyone submitting the website form was dropped into a “tell us your company” sequence. The enrollment filter was an OR across four unknown fields, and the forms only collected an email, so every single submitter matched. Rebuilt to enrich first, then branch on whether a company record actually exists. Buyers who are ready to book now get the meeting ask immediately.
Security & hardware
Leads that nobody had accepted yet
Qualified leads were reaching sales and then sitting, with no way to tell how long. Built a handoff SLA engine that measures the time between a lead qualifying and a rep accepting it, and fires alerts when that window is missed. Paired it with decay reporting that recycles stalled opportunities back into play with a notification, instead of letting them age quietly inside the forecast.
B2B healthtech
A signed contract that never reached onboarding
A customer signed and the provisioning workflow never fired, because properties it depended on had been archived months earlier. Found, fixed, and retested, then automated the post-signature setup so it no longer depends on someone remembering to do it by hand.
Legal tech
A forecast leadership could defend
Built weighted pipeline reporting and a segmentation model across four firm tiers, plus key-contact tracking that separated primary relationships from secondary ones. Added campaign tracking and attribution so marketing spend could finally be tied to closed revenue rather than argued about.
What I run my own business on
I don't just recommend AI. I operate on it every day.
An AI chief of staff
Custom Claude skills that open and close my day. One pulls today's calls, unanswered client messages across every inbox and Slack, and who is blocked on whom. The other turns the day's requests into tickets and sets chase dates.
Claude Code, Gmail, Slack, Notion, Sybill
Time tracking that logs itself
When client work finishes, Claude logs effort-based entries to Toggl automatically. No timers, no end-of-week guessing.
Claude Code, Toggl API
Invoicing with no manual entry
Toggl hours flow into a Google Sheet, and Apps Script produces branded PDF invoices per client. Nobody retypes an hour.
Google Sheets, Apps Script, Toggl
My own MCP server
Wrote a Model Context Protocol server so Claude can read and manage Google Tasks directly. When a connector doesn't exist, I build it.
Node.js, MCP, Google Tasks API
Want something like this for your team?
Book a free consultation