I’m Georgios Tsotsos, and most people call me George. I’ve spent a bit over twenty years building and leading software and product engineering organizations, mostly around cloud-native platforms, edge systems, and the management planes that connect the two.

The through-line is that I’ve usually been accountable for both halves of a decision, the architecture and the business consequence of it. That is an uncomfortable place to sit, and it is also where the interesting problems are. Almost everything on this site comes out of that position.

What I Work On

Software products and the teams that ship them. For most of my career I’ve led the application side as well as the platforms underneath it. In practice that means product teams across several countries on one delivery cadence, and the practices that keep code quality from degrading as the headcount grows.

Machine learning and agentic AI. ML systems in production, on log analysis at scale and on pricing, and the organizations that built them. Now the work is agentic: control planes on MCP and the security model around them, what an agent may call, what it may write, how that is evidenced afterwards. I lead the teams building them, and some of the architecture decisions are mine. The question underneath is what happens to an engineering organization when a large share of its code is generated rather than written, and who is accountable when it fails.

Cloud-native platforms and what they cost. Multi-region SaaS, control planes, and the operating economics underneath them. I’ve led platform organizations through the phase where the infrastructure bill stops being a line item somebody in finance owns and becomes an architectural constraint that engineering has to answer for. The largest of those efforts took about a third out of a SaaS business’s cloud cost of goods sold (COGS).

Regulated and sovereign environments. I’ve been through a FedRAMP Moderate authorization on Azure Government, and the controls themselves were rarely the hard part. What consumed more time and effort was producing evidence that changes had been reviewed and approved by someone with the standing to approve them. I’ve also designed a PCI-DSS SAQ-D environment and run the engineering organization that operated it, and worked under SOC 2, HIPAA and GDPR, and recently the EU AI Act. The pattern holds across all of them: the regime is usually reasonable, and the difficulty is whether your systems can produce the evidence as a by-product of running normally.

Edge computing and device fleets. Broadband gateways, CPEs, Wi-Fi access points, set-top boxes. The management planes that keep millions of multi-vendor devices patched, configured and observable across intermittent connectivity, built on open standards like the Broadband Forum’s User Services Platform (TR-369) and open-source baselines like prplOS. Lately that includes containers on the gateway itself, which is what the talks below are about.

What I Write Here

This site is where I work through the engineering and leadership questions I’m dealing with anyway. Four recurring threads:

  • AI architecture. Agentic systems, edge AI, and the operational realities of running inference at scale rather than in a demo.
  • Engineering governance. Accountability, audit evidence, and how organizations adapt when AI changes who makes a decision and who can be asked about it later.
  • Platform engineering. Cloud-native and edge infrastructure, control planes, and the trade-offs that decide whether a system holds up under load.
  • Leadership at scale. How technical decisions reach customer value, regulatory exposure, and organizational outcomes, usually later and less directly than anyone plans for.

I try to keep the arguments defensible and the sources checkable, and I’d rather be corrected than be vague. Views here are my own. No current or previous employer reviews or approves them.

Talks and Standards

I speak at industry and open-source conferences, mostly on edge platforms, broadband architecture and container management at the device level.

These talks were given in a vendor role, for the company I worked for at the time. Each page links the original recording; the write-up is my own.

I follow the work at the Broadband Forum, the prpl Foundation, the Linux Foundation and the CNCF, and I’m a member of the IEEE.

Elsewhere

LinkedIn · GitHub · X · ORCID 0009-0004-2813-952X