I build revenue systems that pay for themselves.

ContextA ~100-person construction-management SaaS running its GTM motion on a fragmented eight-tool stack — Salesforce, Salesloft, Mailchimp, Intercom, Delighted, Gong, ZoomInfo and more — with data reliability and reporting quality paying the price.

What I builtI architected and executed the migration to HubSpot nearly single-handedly: 65K contacts, 16K companies and 2.5K deals moved across 3 core pipelines (3 more added since), ~100 workflows including Rippling-driven automatic user provisioning, 42 integrated apps, and custom API integrations with our production database. Build to UAT in 6 weeks; Salesforce replaced in 2 months; full adoption in 3. HubSpot mandated paid partner onboarding — I converted it into 60 hours of Accelant consulting to pressure-test my build, rather than an out-of-the-box migration. I also redesigned pricing and contract workflows with automated tiered pricing and document generation in PandaDoc.

ResultAnnual tooling spend cut by roughly two-thirds — overshooting the original savings projection by about 30% — delivered in-house at a fraction of full-service implementation quotes, with no ongoing support retainer needed.

ContextOutbound is expensive and repetitive: finding accounts with real buying intent, mapping the committee, and writing relevant multi-channel outreach at scale.

What I builtA Claude-based agent skill that hunts intent signals across Clay, ZoomInfo, HubSpot, Snitcher and open web research, identifies the buying committee, drafts sequenced email, LinkedIn and call outreach — then reviews its own results to push the best-performing approaches. Wired into Salesforge, it now sends email and LinkedIn touches autonomously.

ResultIn the first weeks: 5–7 meetings booked, 126 deals tied to signals the skill surfaced, 268 deals carrying a skill-processed contact, and 381 net-new contacts added from intent signals — on top of 26,430 records enriched, 6,171 emails drafted, and 8,059 call tasks given talk tracks.

Honest caveat: the sales cycle is long, so no closed revenue is attributed yet — and I don't claim wins before they close. I report this system the same way I build forecasts.

ContextA ~400-person fleet-management and fuel-card provider with a ~100-rep sales org. Application processing leaned on manual data entry, and the CRM's lead-feeding process generated duplicates that quietly leaked revenue.

What I builtI integrated a government business database directly into the application flow, cutting processing time by 3 minutes per application. Leading a 2-person Sales Ops team and a 10-person CRM dev team, I drove a full platform overhaul — introducing structured change management with testing, training and staged rollouts to protect adoption. I also reworked the lead-feeding process to eliminate duplicates at the source.

ResultRoughly 1,100 work hours saved annually, acquisition tracking accuracy up 20%, and €100K+ in annual losses averted — with ~20K companies enriched at rollout and ~2K a year flowing through since.

ContextAn LSE-listed derivatives trading platform with a ~70-rep sales floor. Leadership needed to know what performance to expect as lead volume swung month to month — and whether a miss was a rep problem or a market problem.

What I builtA model that took each rep's and each team's monthly performance across multiple metrics, fitted them to logarithmic curves, and predicted outputs from inputs: if the team averaged 1,000 inbound leads and received 1,500 this month, here's what should close. Targets were set from a 13-month moving average of individual performance, which made the conversation specific — "at this lead inflow you should have closed 100 wins and closed 90; what happened?" — then moderated with market conditions and peer feedback before easing targets or pushing harder.

ResultAdopted for global C-suite sales forecasting and target setting. Built while I was still an analyst, the model stayed in production for more than five years after I left, until the sales organization was restructured in 2025.