Cloud or on-prem. GPU or inference API. Which vendor, which model, which data egress path. I spent four years inside Meta's data centers and a year at AWS networks. Now I help mid-market IT teams make those calls without the Big-4 deck — and without the regret.
Hyperscale infrastructure thinking. Mid-market scale.
CIOs, Heads of IT, and Heads of Infrastructure at companies between $20M and $500M in revenue. Your team is being asked to support AI workloads. The vendors are calling. The cloud bills are climbing. The choices in front of you — cloud vs. on-prem, GPU strategy, vendor lock-in, data residency — are the kind that don't reverse cleanly.
"The capabilities seem to change from week to week. We're being asked for a strategy on a target that won't sit still."
— CIO quoted in CIO.com, 2025Fixed scope. Real prices. No surprise change orders.
3 weeks · fixed scope
An honest read of where your infrastructure actually stands for AI workloads. Cloud posture, data flow, GPU/inference readiness, vendor exposure, FinOps reality. Ends with a roadmap your CFO will approve and your security team will sign.
8-16 weeks · one workload, to production
We pick the one AI workload with the clearest ROI and stand up the infrastructure to run it well. Inference platform, RAG pipeline, GPU cluster, hybrid edge setup — whichever the assessment said to ship first.
6-18 months · ongoing leadership
Embedded infrastructure leadership for your AI buildout. I sit in your IT leadership cadence, run vendor evaluation, own the architecture decisions, and shield your team from hype. You get a hyperscale-trained operator without a hyperscale hire.
I'm not a Big-4 partner. I don't have an army of associates. I have fifteen years operating inside enterprise infrastructure — networks, data centers, support, capacity — including four years at Meta and one at AWS. I have a strong opinion about which infrastructure choices actually scale, and which ones look great in a vendor demo and fall apart in production.
If a roadmap can't produce a visible win inside a quarter, it's a research project, not a transformation. Every engagement is designed around a shippable artifact, not a deck.
Most mid-market AI infrastructure wins come from picking the right managed service and integrating it properly — not from running your own GPU cluster. I'll tell you when to self-host, and the answer is "rarely."
Hyperscalers oversize because they have to. You don't. The capacity decisions that make sense at Meta scale will bankrupt you at mid-market scale — and vice versa. The math is the work.
I'll be as clear about what to kill as what to ship. Half the value of an outside operator is the permission to ignore the 15 infrastructure vendors who emailed you last week.
Fifteen years operating inside enterprise infrastructure. Four years at Meta in datacenter capacity engineering. One year at AWS as a Network Technician II. Now independent.
I spent the last four years inside Meta's data centers planning capacity for hyperscale infrastructure, and a year before that building networks at AWS. Fifteen years total inside enterprise IT, from network technician through hyperscale operator.
That perspective is the asset you're hiring. Most consultants selling AI infrastructure to mid-market IT teams have never planned capacity inside a hyperscaler. They've read about it. I've done it.
I started this practice because almost every mid-market IT leader I talk to has the same problem: they're being asked to support AI workloads, and the advice they're getting is either Big-4 decks priced at $500K+ or LinkedIn influencer slop. There's no honest middle.
I bring the infrastructure thinking I learned at Meta and AWS to teams that need it without the hyperscaler budget. Charlotte, NC. Serving clients across the US, remote with periodic on-site.
Tactical notes from inside the work — the wins, the things I killed, and the vendors I'd never touch again.
Short, tactical posts on the AI infrastructure choices that actually scale in mid-market environments — and the ones that look great in a vendor demo and fall apart in production.
Follow on LinkedIn →A weekly note for mid-market IT leaders — one tactical piece, three news items with my take, one tool I'm using, one play to try this week.
Subscribe at TMC →Tell me where your infrastructure is and what AI workloads you're trying to support. I'll tell you the three things I'd do first, and whether you actually need me or just need a nudge in the right direction.
Book an infrastructure callPrefer email? hello@yasinjabbar.com