A narrow specialism and a bias against complexity.
Quarkray is a one-person engineering practice. Not an agency with one visible engineer, not a network of subcontractors — one senior engineer who scopes the work, writes it, and hands it over.
01 The practice
What that means in practice
The specialism is big data engineering: platforms that hold real volume without a team of five keeping them alive. The range is broader — APIs, applications, cloud infrastructure — because data platforms don't ship in isolation and the seams between vendors are where projects fail.
Working solo is a deliberate structure, not a stage before hiring. It removes the layer that costs clients the most: the gap between the person who sold the work and the person doing it. You explain the problem once, to the engineer who will fix it.
The trade-off is capacity, and it's stated plainly on the how it works page rather than discovered in month two.
| Structure | Independent, one senior engineer. No subcontracting without telling you. |
|---|---|
| Focus | Big data engineering — platforms, pipelines, streaming, lakehouse. |
| Also builds | APIs, internal tools, front-ends, cloud infrastructure. |
| Engagements | Architecture review · scoped build · monthly retainer. |
| Working style | Remote-first, European hours, asynchronous by default. |
| Contact | contact@quarkray.ro |
02 Principles
Six opinions that shape every engagement.
These aren't values-page filler — they're the things that actually change what gets built, and the reason some proposals recommend doing less than you asked for.
Boring technology, deliberately
Every novel component in a platform is a thing your team will have to learn, operate and eventually debug at 3am. New tools have to earn their place against the one you already run.
Open formats, always
Open table formats and infrastructure as code keep the exit door unlocked. If a platform can't be moved, its vendor sets your prices — and eventually does.
Measure before optimising
Most "slow pipeline" tickets are one skewed join or one full scan. Profiling first regularly turns a rewrite into an afternoon.
Write the decision down
A short record of what was chosen, what was rejected and why. Six months later that document is worth more than the code it explains.
Delete more than you add
Unused dashboards, dead tables, pipelines nobody reads. Removing them is often the highest-return work available and almost never on the roadmap.
Say no in writing
If the thing you asked for won't solve the problem you described, you'll get that in the proposal — with the cheaper alternative next to it.
03 Fit
- A scale-up whose data outgrew the setup that got them here
- A team with engineers but no one who has built a platform before
- A CTO who wants a neutral opinion before a large commitment
- A product with an AI feature that needs real data infrastructure behind it
- A company that wants to own what gets built, not rent it
- You need four engineers starting next month — that's a hiring plan
- You want a body to fill a seat rather than a problem solved
- The decision has been made and you need someone to validate it
- The work is a licence resale dressed as a consulting engagement
- Nobody internally will have time to receive the handover
If you're in the second column, the call is still free and you'll get a straight recommendation — often for a different kind of help entirely.
04 The name
Quarkray
A quark is the smallest thing that still counts as a constituent — you can't isolate one, it only exists bound to others. A ray is direction: a single origin and a path outward.
Which is a reasonably accurate description of a data platform. Individual records mean nothing alone; the value is entirely in how they're bound together and where they're pointed. The mark is an orbit around a nucleus, for the same reason.
Next step
Find out if this is a fit.
Thirty minutes, no charge. If the answer is that you need something other than an independent data engineer, you'll hear that on the call.