Technical work
Practical work with infrastructure and AI
My technical foundation remains active through personal projects and day-to-day use of infrastructure, cloud services and AI tools. I use that work to understand the decisions behind the technology: access, data flow, dependencies, cost and recovery.
These are personal projects developed with AI assistance. I define goals and constraints, use AI to help implement and troubleshoot, and review the resulting behavior. This is practical technical experience alongside my sales career.
Private Infrastructure Operations Platform
Personal project · AI-assisted development
Problem
I wanted an external view of useful operational information without giving that view authority to control the private infrastructure.
Design
The documented design keeps authoritative state locally. An allowlisted snapshot contains only approved information; authenticated publication sends it to a Cloudflare Worker and D1 bridge. A private external view reads that information through a server-side connection.
The distinction between current, stale and unavailable evidence matters. Missing information must not be mistaken for a healthy system. The external view has no infrastructure-control path.
My contribution
I defined the purpose and information boundaries, directed AI-assisted implementation, and reviewed the behavior against those requirements. The project gives me a concrete way to think about the difference between showing information and granting authority.
Sharing operational evidence while keeping control private
- 1 Private operational state Authoritative records remain in the private environment.
- 2 Selected snapshot Only an allowlisted subset is prepared for publication.
- 3 Authenticated publication A signed request carries the selected snapshot.
Selected evidence only
- 4 Worker and D1 The external application receives and stores the published evidence.
- 5 Authenticated read view A server-side check gates access to the published view.
- No infrastructure control path from the external view
- Missing or stale evidence is not a healthy result.
Operating the environment behind the project
My personal environment includes Linux, Proxmox virtual machines and containers, networking, storage, backups and services. I use AI assistance for commands and implementation, while working through dependencies and the consequences of changes.
The useful questions are practical: what depends on this service, what evidence shows that a change worked, and what would be needed to recover it? Git and GitHub help me review changes and retain a route back to earlier versions.
I also use Cloudflare Pages and Wrangler for personal website deployment. A separate private portal uses Cloudflare Access; it is distinct from the operations platform’s external view.
Working with local and hosted AI
I use OpenClaw for persistent agent workflows, Ollama for local model inference, and Whisper for speech recognition. Hosted models provide another option when a task needs different capabilities.
The choice involves more than model quality. I consider what information can leave the local environment, the available computing resources, latency and cost. Local transcription does not mean that every later stage of a workflow stays local.
AI assists the implementation. My responsibility is to define what the workflow should do, review its output, and decide whether the evidence supports the result.
Why this matters in a customer conversation
Practical work helps me ask more useful questions about a customer’s environment and recognize when a specialist needs to be involved. I can connect a commercial discussion with the technical conditions that affect evaluation, adoption and ongoing use.