Who we are
Boutique is the strategy, not the stage.
Pink Cloud LTD is a small consultancy working where AI systems, cloud infrastructure, and security meet. We are not trying to become a large firm. The people who scope your engagement are the people who deliver it, and that stops being true above a certain headcount.
How we work
A short list of things we will not do.
Most consultancy positioning is a list of strengths. Ours is more useful stated as constraints, because constraints are checkable and adjectives are not.
No percentage of savings
Fixed fee. There is no upside for us in recommending a three-year commitment you should not sign.
No reseller margin
We hold no partner margin on any platform we might recommend, so the recommendation is just a recommendation.
No findings held back
Everything we find goes in the report at the time we find it, not into the scoping document for a follow-on engagement.
No lock-in by design
Deliverables include the harness, the code, and the reasoning. If you never call us again, everything still works.
No client named without permission
Case studies are anonymized by default. Written approval is required before any engagement appears in public, in any form.
No work we cannot staff properly
When we are full, we say so and offer a date. We do not backfill an engagement with people who were not in the scoping call.
The team
We present as a collective. Here is what is in the room.
Offensive security work frequently cannot carry a byline. Some of it sits under client NDAs that outlast the engagement; some of it comes from backgrounds with obligations that outlast the posting. Publishing headshots and job titles would mean either thinning what we can say about the work, or saying it about people who cannot be named alongside it.
So the capabilities are public and the roster is not. Before you sign anything, you are told under NDA exactly who is staffing your engagement, with the specific background each person brings to it. That is a firmer commitment than a team page, which in this industry is frequently a list of people you will never actually meet.
- Signals intelligence
A Unit 8200 alumnus
Nation-state tradecraft applied to commercial systems. The discipline of treating every input channel as hostile long predates the LLM era, and it transfers to this work almost without translation.
- Adversarial
Offensive AI red-teamers
Practitioners whose working week is spent finding routes through model guardrails — prompt injection, tool-use abuse, and the chained failures that single-shot testing never surfaces.
- Economics
FinOps specialists
Cloud economics at the resolution that matters: unit cost per tenant, per request, per thousand tokens — including the inference workloads where cost and abuse turn out to be the same problem.
- Measurement
Model-evaluation authorities
Depth in how model behaviour is actually measured, and in building the harnesses that make alignment and capability claims checkable rather than asserted.
The name
A pink cloud is the calm right after a storm, and — in recovery circles — the early stretch where everything feels solved. Both readings are useful in this line of work. Systems feel safest immediately after the incident that proved they were not, and confidence about an AI deployment tends to peak precisely when the least has been measured.
We kept the name because the reminder is a good one. The work is finding out whether the calm is real.