SD Tech Solutions is a seasoned IT shop covering the full stack — physical
cabling and server racks at one end, on-prem and cloud-hosted clustered
compute in the middle, custom software and AI on top.
The work spans IT admin, IT ops, DevOps, software development, and SaaS
delivery for hire. Engagements range from a single-site office build-out to
multi-site networks, hybrid cloud workloads, and on-prem private LLMs that
replace per-seat AI subscriptions.
The goal across all of it is the same: environments that are understandable,
repeatable, and easy to support — built once, documented properly, and
handed back to the team in a state where they own it.
Focus Areas
End-to-end IT for small and mid-size organizations: physical
infrastructure, identity, networking, storage, cloud, software, and AI.
Physical infrastructure I personally oversee or run: ethernet cabling, server-rack builds, structured cabling, UPS, and patch panels.
Networking from the ground up: L2/L3 switching, VLAN design, hardware firewalls, site-to-site VPNs, and wireless point-to-point bridges and backhauls (CPE710-class radios for distance links).
On-prem servers — Windows Server (AD DS, DNS/DHCP, GPO, IIS, Hyper-V) and Linux (Ubuntu, RHEL/Fedora) — alongside dedicated backup servers and snapshot-based recovery.
Cloud and hybrid: Azure (VMs, Entra ID, Intune, M365, Purview), AWS (EC2, S3, IAM, networking), and clustered VM workloads that span on-prem and cloud.
Software for hire: PHP/Linux web apps, custom internal tools, ticketing portals, dashboards, and SaaS-style products billed by retainer or fixed scope.
ERP and CRM integration: pulling enterprise data into custom workflows, automations, and AI assistants so the systems already in place actually talk to each other.
AI as part of the stack — not bolted on: on-prem private LLMs for sensitive data, agentic workflows (Claude Code, MCP, custom orchestrators), and AI-assisted ops.
Approach
Assess what’s actually there, stabilize what’s fragile, then modernize and
automate — including AI where it pays for itself.
Start with a real assessment: walk the rack, read the firewall, audit AD/Entra, look at the backups — not just a questionnaire.
Stabilize the fundamentals first — identity, networking, storage, and backup — so the daily fire-fighting stops before any redesign work begins.
Modernize pragmatically: replace what genuinely needs replacing, virtualize what makes sense, move to cloud only where the math works.
Automate the repetitive: PowerShell and Bash for ops, Power Automate for business workflows, IaC-style configs for repeatable rebuilds.
Layer in security as a default — hardened baselines, MFA everywhere, least-privilege access, central logging — not as a separate "security project."
Bring AI in where the ROI is real: a private on-prem LLM running on a single capable server can replace dozens of $200/seat subscriptions for many orgs, with full data sovereignty.
Document as I go: diagrams, runbooks, and a maintenance plan an in-house team can actually pick up.
How I Work
Clear scope, small proofs, hands-on delivery, and documentation that
outlives the engagement.
Define scope, success criteria, and constraints up front so everyone knows what "done" looks like — and what is explicitly out of scope.
Run small proof-of-concepts before rolling changes across an environment: stand up the new firewall in parallel, mirror the workload, validate, then cut over.
Engage hands-on: I run the cabling, rack the server, configure the firewall, build the AD/Entra design, write the script, ship the app — and supervise the trades I don’t do myself.
Use git, configuration-as-code, and versioned scripts so changes are reviewable instead of living in someone’s head.
Bring creative answers when the constraint is real — older systems, tight budgets, regulated data, or no internet at the destination.
Iterate: ship something useful, gather feedback, refine. Avoid month-long radio silence followed by a big-bang reveal.
Hand back diagrams, runbooks, and a maintenance schedule the in-house team can follow on a busy Monday morning.
