Case study · Fractional CIO/CTO
Metrix Group: a year-long engagement that ended well
The hardest thing to prove about fractional work isn't that it starts well — it's that itends well. Here is one that ran its full course: I built a technology function from zero, and a year later handed a running platform and team to a full-time CTO when the company was acquired. The client is on the record about it.
- Role
- Fractional CIO/CTO · ~15 hrs/week
- Duration
- About a year, 2025 → 2026 · completed
- Funding
- DMAP approved (Ontario Centre of Innovation) + SR&ED
- Team
- Hired and led a 6-person cross-functional team
- Platform
- Neo4j GraphRAG + MCP server on Kubernetes
- Shipped
- 18 AI workflows identified · 6+ in production
- Exit
- Acquisition → handed off to a full-time CTO
What I walked into
Metrix Group has deep domain expertise in pharma behaviour-change and sales training, and wanted to become an AI-enabled platform business rather than a traditional services firm. What it didn't have was anything to build that on: no CTO or technical function, no in-house developers, no cloud infrastructure, and a regulated-industry constraint that makes most shortcuts illegal — pharma clients require strict per-client data isolation. There was also a government funding window open, but it was unclaimable without a credible technical strategy to put behind it. I came in as fractional CIO/CTO to own all of that end to end.
Securing the runway
The first real win wasn't code — it was funding. I led the DMAP (Digital Modernization and Adoption Program) submission to the Ontario Centre of Innovation end to end: process-mapping every business area, then authoring the assessment, the platform documentation, and the data-architecture schematic that made the case. From process mapping in the late summer to approval the following January, it secured non-dilutive government funding and, just as importantly, validated the strategy before we spent a dollar building. I opened a second non-dilutive channel by structuring the SR&ED R&D tax-credit claims for the AI work.
The platform
At the core is a proprietary AI platform the company owns outright, not a seat rented from a vendor. A Neo4j knowledge graph is the analytics backbone — it models behaviour-change frameworks, stakeholder gaps, and evidence chains as a graph — on an ontology-first GraphRAG architecture, with a Model Context Protocol server (Python and FastMCP, ~20 tools) on Kubernetes exposing it to AI agents, and per-client isolation enforced at every layer for pharma compliance — a separation I had validated by legal counsel, not just asserted. It runs on infrastructure I built from zero: an Azure AKS production cluster, AWS for tenant isolation, a Bronze/Silver/Gold data lake fed by Airbyte, and a full Microsoft 365 migration — on a stack of self-hosted n8n, Azure OpenAI, Neo4j, Qdrant, and MongoDB. (Why a knowledge graph was the load-bearing decision is a story on its own —why the fix was context, not a bigger model.)
What actually shipped
We identified eighteen candidate AI workflows and ran a deploy-and-measure discipline against them rather than boiling the ocean — more than six reached production by the end of the engagement. The ones that moved the needle:
- A pipeline that turns interview transcripts into structured behaviour-mapping tables through the knowledge graph — cutting analyst time on those reports by 50–70%.
- AI-generated workshop materials — facilitator guides, activities, slides — that take workshop creation from days to hours.
- An AI sales-training builder that assembles tailored learning paths, deployed for a global pharmaceutical client.
- Augmented-reality training characters, shipped to production with reusable deployment tooling.
- A learning-impact analytics dashboard for measuring training outcomes.
The team behind it
A platform is only as durable as the people who run it. I hired an AI developer and led a six-person cross-functional team — developers, designers, and data specialists — on daily standups, Scrum, and a full SDLC, managing external suppliers alongside them. Much of the quiet work was de-risking: disaster-recovery planning, security controls, the legally validated pharma data separation, and knowledge-transfer plans specifically to reduce the bus-factor risk that kills small technical teams.
How it ended
The engagement concluded when Metrix Group was acquired and brought on a full-time CTO. I handed over the platform, the team, and the processes, and they kept running without a disruption. A fractional engagement is meant to make itself unnecessary — this one built a technology function durable enough to outlast it, and handed off clean to the person who took it forward.
“I'm impressed by Reyem Technologies Inc's broad expertise, AI fluency, and positive, supportive leader.”