If you work in hospital IT, or want to, there is a role forming quietly between the security team and the data team. It has no settled job title, and the shorthand people use for the skills is SecDataOps. The job is to make sure the logs that security depends on are complete, trustworthy and affordable. This page lays out what the work involves, which skills carry over from adjacent roles, and a realistic plan for the first ninety days. It covers administration and technology only; there is nothing clinical here and nothing that should be read as advice.
What the role actually does
The Ops Atlas calls the discipline “data engineering for defenders”. In practice that means bringing each source of events into the pipeline, parsing it into a common schema, enriching it with useful context, routing it to the right storage tier and keeping it for the right amount of time. The aim is that when an investigator or auditor later needs evidence, it is there, readable and not ruinously expensive.
It sits next to other jobs but is not the same as them. Writing detection rules, triaging alerts and containing incidents are different roles. The person in this seat looks after the evidence supply chain, from producers to collection to pipeline to storage tiers to the people who consume the data.
Skills that transfer
- Systems administration: knowing how servers, domain controllers and network devices emit logs, and what changes when they are upgraded.
- Data engineering habits: schemas, fixtures, tests, pipelines as code, and the discipline of version control.
- Scripting and regular expressions: parsers are small programs, and sloppy ones cause quiet damage.
- Cost thinking: being able to say what a stream costs per day and what happens if you re-tier it.
- Plain writing: contracts and runbooks only work if the next person can read them.
Skills worth learning on purpose
Common schemas
Different tools describe the same event in different dialects. Open schemas such as OCSF try to give everyone one vocabulary, and the Atlas notes that rival schemas are converging. You do not need to memorise either. You do need to be able to read a mapping and spot a field that was lost in translation.
Retention and its sources
Retention is where policy meets storage. The Atlas mentions federal logging rules in the United States that split logs into maturity tiers and expect a year of readily searchable data plus eighteen more months in colder storage, a controls catalogue from NIST that covers audit record retention, and a 180-day rule in India. Being able to cite where a number comes from will make you useful in meetings.
Failure modes
Good practitioners collect small, specific horror stories. One parser converts an empty value to zero, which means junk text can masquerade as a refused connection to port zero; the cure is a strict pattern that sends unreadable lines to quarantine. Another buffers events in memory only, which is not durable, so a disk buffer is wiser. Old-style syslog cannot acknowledge delivery. Replays double-count unless you deduplicate on a stable identifier. These details separate a candidate who has read about the field from one who has operated in it.
A ninety-day plan for the first quarter
The Atlas offers its own thirty, sixty, ninety outline, and it translates well into a new-job plan:
- Days 1 to 30: build the coverage register, score every source, set retention on purpose and add health monitors to the ten most important sources.
- Days 31 to 60: write a contract for each of those ten, add test fixtures and automated checks for their parsers, plant canaries and put a price on each stream.
- Days 61 to 90: shift the three heaviest feeds into cheaper tiers, replay seven days of raw data to prove every alert rule still fires, then hold your first silence exercise and write up a baseline.
Another habit worth building is writing down the assumptions behind each decision. When you shorten retention on a noisy stream, record what you checked first, who agreed and when it should be reviewed. Months later, an auditor or a successor will thank you, and you will be able to defend the choice without relying on memory. Pair that with a calm, steady tone in change discussions, because people guard their logs more fiercely than you would expect.
Where to read more
For definitions and the maturity ladder, start with the Ops Atlas page for SecDataOps. It describes five stages of maturity, starting at reactive and ending at optimising, with inventoried, contracted and measured between, which gives you a vocabulary for interviews and for your own progress reviews.
If you are weighing the wider ecosystem of operations careers, our overview of modern IT operations disciplines and how people learn them is a handy companion, and the piece on building a foundation in software delivery and reliability work covers the engineering base that this role builds on.
Practical advice for the interview
Bring a one-page sample. A coverage register with a handful of invented sources, scored on the five dimensions, shows more than a long list of tool names. Be ready to explain one failure mode in your own words and to estimate the daily cost of a stream. Do not claim certifications that do not exist for this label; the Atlas is clear that no standards body defines it, and honest candidates stand out.
Finally, remember where you work. Hospital environments are sensitive, and your job is to protect evidence about systems, not to touch anything clinical. Keep your scope narrow, document what you change and ask before altering any feed that other teams depend on. This article is general career information, not legal, medical or compliance advice.
