I build systems. Some run on code. Some run on stories.

On-time delivery sat near 60%. Every status meeting had the same shape: a team confident on Monday, slipping by Thursday, and a stakeholder who’d stopped believing the dates. The ambiguity wasn’t who was at fault — it was that no one could see where work stood until it was already late. I resisted adding process to teams already drowning in it. Instead I made execution visible: clear scope boundaries, prioritization tied to business outcomes, one honest status cadence. The tradeoff was uncomfortable — we committed to less, out loud, and held the line. Within six months the dates became unremarkable, which is exactly what a high-growth org needs them to be.

At Clifford Chance I owned a document-management migration across ten international offices. The constraint wasn’t the schedule or the £5M budget — it was that a global law firm cannot pause. A document unavailable for an hour in London is a deal team that doesn’t stop for our migration. So I planned for zero operational downtime as a hard floor, not a target. We sequenced by office and time zone, rehearsed every cutover, kept rollback ready at each step, and held governance in the room the whole way. We landed all ten offices with no operational downtime and a full audit record. In a regulated environment, the unglamorous work — the rehearsal, the rollback plan — is the work.

At Builder.ai’s Studio Store I ran delivery across 40+ concurrent enterprise engagements spanning the US, UK and APAC. Some were already at risk when they reached me, and the at-risk accounts rarely announced themselves. The loud ones I could see. The quiet ones — clients who’d stopped escalating because they’d stopped expecting a fix — were the real exposure. So I treated recovery as a communication problem before a delivery one: I went back to the accounts that had gone silent, named the gap honestly, and rebuilt the cadence. Predictability is what wins enterprise trust back. We recovered the at-risk accounts and held renewals above 85% across the portfolio.

Incidents were climbing, and the same few engineers kept getting paged because the critical paths lived in their heads. I’d sat through those incident bridges and signed off the recovery the next morning, so I knew the real risk wasn’t the outage — it was the key-person dependency underneath it. The hard call was to slow merge velocity while we built reliability in: observability that surfaced problems early, blameless post-mortems with real follow-through, rotated on-call, cross-training, documented ownership. Some of it felt invisible against shipping pressure, and I had to hold leadership’s confidence while it compounded. Reliability stopped depending on who was awake — and the late-night calls went quiet.