Sales Engineer
Alistair
George
Commercial instinct.
Technical execution.
The true solution comes from asking the right people the right questions.
LinkedIn
Sales Engineer
Commercial instinct.
Technical execution.
The true solution comes from asking the right people the right questions.
LinkedIn
The arc
The numbers game became a coaching game. Built a discovery framework that spread company-wide, then moved into account management running a sell-with-me motion — analysing partner books, building product strategies, coaching smaller agencies on growing their clients' businesses. Partners treated me like a member of their team because I operated like one.
"Why is the platform doing X?" replaced "Can it do X?" That shift — from feature presenter to problem investigator — is where the architecture instinct surfaced. Managed high-complexity global ISV partnerships worth $2M+ ARR. Learned that the real answers surface when you ask the right questions of the right people — not the first people who respond.
A high-velocity ISV partner disputed invoices persistently. Balance reached ~$250k. Product threw up their hands. Reverse-engineered the billing system using BigQuery and LLM-assisted analysis — no internal tooling access. Found the edge case: SMB products cancelled in a slim pre-cutoff window, billed at scale. Presented the evidence and a settlement proposal. Partner agreed. Built a credit process and a daily forecasting query. Not a technical job. A solutions problem.
In late 2024, pitched the acting CRO directly on creating a new role — one page of notes, the rest verbal. Six months later, a Neighborly crisis became the catalyst that made it real. The commercial element went to dedicated AMs. The technical element stayed. The role didn't change those relationships. It deepened them.
How I work
Diagnostic before prescriptive
Customers know what they're asking for — not always what they need. The job starts with the right questions to the right people: surfacing the real pain, the unforeseen constraint, the problem behind the stated problem. Only then does the solution take shape.
Solutions architecture done right
Anyone can build something that works. The question is whether it's right for the problem — asking questions before writing code, understanding the full system, pushing back when the spec is wrong, choosing the right tool for the task. "It works" and "it's good" are not the same thing.
Revenue outcomes
Real, quantifiable impact on the customer's business — not vanity metrics, not "we shipped a thing." The measure is: did it move something that mattered for them?
Extreme ownership
Projects don't languish. If something has my name on it or sits in my orbit, I'm invested in its success. Personal standards, not professional performance.
AI at the orchestration layer
Not a user of AI — a builder with it. Multi-agent workflows, context engineering, tool design. Builds internal tooling that other people then rely on. Uses the right model for the task: Claude for reasoning, lighter models for throughput, local models where appropriate.
Work
US home services franchise · 3,500+ locations · 75k conversations/month
Neighborly Conversations Analysis & ROI Report
A franchise network was paying millions for AI assistant deployment across franchise owner businesses — with no reporting to quantify ROI on that spend. Built a three-layer classification pipeline: deterministic rules for speed and confidence, ML for mid-confidence records, LLM resolution for edge cases. Scheduled Cloud Run application, BigQuery pipeline, React UI with Firestore-indexed filtering and visualisations delivered to the customer.
Built for a customer. Became the architectural foundation for the next build.
Internal · Vendasta CS team
Vector AI Conversations Reporting
Vendasta's customer-facing AI assistant was live across the platform with zero visibility into where it was winning, where it was failing, or what conversations looked like at scale. Built a reporting system on the same architectural foundation — added one component: an SLM transformer to partition long conversation streams into discrete discussions before classification. The existing classifier stack pointed at the new records.
Almost no new architecture required. This is what reuse looks like when the first thing was built right.
ItaliaOnline · Platform UI · tens of thousands of strings
AI-Driven Platform Localisation Pipeline
Started with an urgent request from a major Italian digital media platform: audit and update the Italian strings across their instance. Delivered 3,600 updated strings across 24,000 audited — under 24 hours, running as a background task. Built the Weblate MCP and Claude integration that made it possible, then extended the architecture to cover the full platform UI. Context-aware translation informed by a platform-specific canonical glossary, not just raw machine output.
Two hours of native-speaker time required. One for the canonical glossary at kickoff. One to review low-confidence translations at close.
ISV partnership · high-velocity domain registrar
The Billing Investigation
Persistent invoice dispute. Partner claimed overbilling. Balance at ~$250k. No internal tooling access — reverse-engineered the billing system with BigQuery and LLM-assisted analysis. Found an edge case that only surfaced at high SMB velocity: products cancelled in a slim pre-cutoff window, billed at scale. Presented the evidence and a settlement proposal. Partner agreed. Built a manual credit process and a daily forecasting query.
The origin of everything: not "can we build it" — but "why is it doing this, how do we stop it, how do we see it coming."