Careers

Four open roles in clinical data, statistics and agent reliability

We are a small team doing detailed work for regulated clients. The listings below are the actual work, written by the people who do it, not a job description generator.

Open positions

Every role here is open and funded

When one is filled we take it down. We do not keep listings up to collect CVs.

Clinical data and ADaM

Clinical Data Engineer, SDTM and ADaM

Build the pipelines that take collected study data to compliant SDTM and ADaM, with traceability that holds up when an inspector follows a value back to its source record.

What you would do

Write metadata driven pipelines so the same study rebuilds identically every time it is run.
Produce define.xml that documents the data rather than describing a wish.
Maintain traceability from CRF through SDTM and ADaM to the analysis table.
Review other people's specifications and say when a derivation will not survive inspection.

What we need to see

SAS or R for clinical programming, and enough of the other to read it.
Working knowledge of CDISC SDTM and ADaM, not just awareness that they exist.
Two years or more producing datasets that went to a sponsor or a regulator.
The habit of writing down why a derivation is what it is.

Helps, not required. You have been on the receiving end of a define.xml validation report and fixed it.

Apply for this role

Biostatistics and estimands

Biostatistician, Estimands and SAP

Turn a clinical question into an estimand a statistician and a clinician both recognise, and write the analysis plan that follows from it.

What you would do

Draft and review statistical analysis plans, including the estimand framing and intercurrent event strategy.
Design missing data handling as a decision made in advance rather than a patch applied later.
Specify and check analyses, and challenge results that look better than they should.
Work with programming so the SAP and the code say the same thing.

What we need to see

MSc or PhD in statistics, biostatistics or equivalent working depth.
Practical familiarity with ICH E9(R1) and what an estimand actually commits you to.
R for analysis, and the ability to read someone else's SAS.
Three years or more in clinical trials, industry or CRO.

Helps, not required. You have defended an analysis choice to someone who disagreed with it.

Apply for this role

R validation and submissions

R Validation and Submissions Engineer

Stand up and qualify the statistical computing environments that regulated analysis runs in, and get R based work through a submission without it becoming the story.

What you would do

Qualify R and Python platforms with documented evidence: package qualification, audit trails, access control.
Build the reproducibility guarantees that make a result defensible two years later.
Prepare and package analysis for regulatory submission, including the parts reviewers ask about.
Keep 21 CFR Part 11 readiness as a property of the system rather than a document about it.

What we need to see

Strong R, including package development and dependency management.
Experience with validated or qualified environments in a regulated setting.
Comfort with containers, CI and infrastructure as code.
Two years or more in a GxP or comparable regulated computing context.

Helps, not required. You have contributed to pharmaverse or a comparable open source clinical R project.

Apply for this role

AI agents and reliability

AI Agent Reliability Engineer

Measure whether an agent actually does what it claims, build the benchmarks that show it, and publish results that survive someone trying to reproduce them.

What you would do

Design and run reliability benchmarks for AI agents on real tasks, not curated demos.
Build harnesses that make a failure reproducible rather than anecdotal.
Write up findings honestly, including the runs where our own tooling came off worse.
Turn benchmark work into product signal for our agent tooling.

What we need to see

Strong Python, and R or a willingness to learn it properly.
Experience evaluating models or agents, including where standard metrics mislead.
The instinct to distrust a number that is too good.
Two years or more in ML engineering, evaluation or a research adjacent role.

Helps, not required. You have published a benchmark someone else has criticised.

Apply for this role

How we hire

Four steps, about two weeks

You send something real

A CV works. A repository, a define.xml you fixed, or a write up of an analysis that changed under scrutiny tells us far more.

A conversation about your work

Forty five minutes on what you have built and where it broke. No algorithm puzzles.

A problem from our backlog

Timeboxed, paid, small enough for an evening. The work stays yours.

Offer or a clear no

You get an answer with a reason either way. Nobody is left waiting.

Not ready to apply

If you are still building toward these roles

Our industry programme runs the same four tracks over eight weeks, with review from the engineers who do this work for clients and a certificate issued by EDUSHARK TRAINING.

It is paid training rather than employment, and we are explicit about that on the page.

Where our people come from

Clinical programmingSAS backgrounds moving into R.
Academic statisticsWanting applied trial work.
Software engineeringCurious about regulated environments.

Think one of these is you?

Write to careers@techworkslab.com and say which role and what you have built.

Apply now