Software Engineer

Amplifier AI is building a surgical planning 
platform powered by our own imaging, 
segmentation and 3-D measurement engine. 
We are a small team of engineers, radiologists 
and data scientists from Ukraine and across 
Europe. You would join the team developing the 
internal Python engine behind 3-D 
reconstruction and measurement for CT-based 
workflows.

Read this before you apply

We hire for reasoning about 3-D space, not 
for a title. The code itself is ordinary clean 
Python — the hard part is everything the 
geometry touches:
Voxel space vs. physical space, and every 
transform between them
Origin, spacing and direction — consistent 
across scanners, series and vendors
Real hospital data: incomplete series, odd 
orientations, inconsistent metadata
Geometric edge cases — surfaces that self-
intersect, clips that miss, distances that are 
"almost" right
Reproducibility: the same input must give the 
same number, today and in six months
If you can explain why a model is offset, 
mirrored or at the wrong scale — not just that it 
looks wrong — you will do well here.
Because the output feeds real surgical planning,
correctness is not negotiable. A wrong 
millimetre is a clinical problem, not a bug report.
This role is probably not for you if you want 
fully predefined tickets, a fixed roadmap, or to 
implement tasks without understanding the 
domain. Hospitals and real clinical data change 
priorities; we adjust as a team.

What We're Looking For

Python engineering (the core of this role)
- Clean, modular, strictly typed Python. We run 
mypy --strict, ruff, pytest, pre-
commit — and deep, technical code reviews.

- Confident with NumPy / SciPy and array-
heavy numerical code.

- Ability to own a feature end-to-end: from 
understanding the clinical problem to 
validating the result on real data.

3-D medical imaging fundamentals
-Understanding of origin, spacing and 
direction, and the difference between voxel 
and physical coordinates.

-Reading DICOM series correctly, validating 
RAS orientation, resampling without silently 
corrupting geometry.

-Comfort debugging complex geometric 
problems: distance, intersection, clipping, 
surface generation from labels.

Toolchain
-SimpleITK, VTK or PyTorch experience is a 
strong plus. If you haven't used them, you 
must be genuinely comfortable diving deep 
and learning independently — we mentor, but 
initiative is expected.
-Docker, Git, GitHub Actions, structured 
logging.

Ways of working
-Clear technical communication — you can 
explain your reasoning and defend a design.
-Ownership over micromanagement. Good 
written and spoken English.

What You'll Do
- DICOM & SimpleITK: read series correctly, 
validate RAS origins and directions, resample 
safely, maintain voxel ↔ physical transforms.
- VTK & geometry: generate surfaces from 
labels; implement distance, intersection and 
clipping algorithms; export reliable STL 
markers and geometry artifacts.
- Measurement & QA logic: build robust 
procedural tools for radiologists and handle 
the edge cases real hospital data produces.
- Pipeline reliability: performance tuning, 
structured logging, and making sure 
pipelines survive messy multi-scanner 
datasets.

Example challenges
- Build a robust distance-measurement tool 
between anatomical segmentations, with 
exportable nearest-point markers.
- Extract centerlines and anatomical landmarks
from noisy CT scans.
- Enforce consistent spacing / origin / direction 
across multi-scanner datasets.

If these sound exciting rather than 
overwhelming, you'll likely enjoy this role.

Nice to Have
- VTK surface / volume rendering experience.
- ML segmentation metrics and tooling (Dice, 
Hausdorff, nnU-Net, PyTorch).
- GPU acceleration and large-volume 
performance work.
- Background in radiology, biomedical 
engineering or clinical imaging.

Our Stack
Python 3.13+ · SimpleITK · VTK · NumPy / SciPy 
· Pydantic · PyTorch · nnU-Net · pytest · mypy 
(strict) · ruff · pre-commit · Docker · GitHub 
Actions · Git LFS · Sentry

What We Offer
Competitive salary, optional stock options, and 
a clear path toward senior-level ownership and 
technical lead responsibilities. Fully remote, 
small team, no corporate theater.

How to Apply
- CV or LinkedIn
- GitHub or portfolio — we value real code
- A short note: why does working in an 
evolving product environment excite you?