Integration Engineer, Clinical Data Path

a2z Radiology AI · Boston (hybrid), US-remote, or remote international · customer-facing, placed to match the account base

The model can be right and the deployment can still fail on the wire. A study arrives incomplete. A sender never closes an association. A multipart parser reads a body into memory and the task disappears behind a bare 502. A firewall rule produces a mid-stream 403 that looks exactly like a TLS failure.

This role owns that wire: how studies move from hospital systems into a2z, how results get back, and how we prove the path is working.

We're building clinical AI that reads alongside radiologists. Our abdomen-pelvis CT triage device is FDA-cleared, and it's the first commercial system to simultaneously triage seven urgent conditions on abdomen-pelvis CT in the U.S. Backed by Khosla Ventures.

Why this role, why now: a cleared device only creates value when it works inside the systems a radiologist already uses. Our clinical data path spans software running inside hospital networks, cloud DICOMweb ingest, de-identification, tenancy, result delivery, and the evidence a hospital IT reviewer expects. Two gateway repositories need continuous integration now. The integration surface needs one owner who follows the study end to end.

What you'd own:

• On-premise DICOM ingest. A pynetdicom C-STORE service that assembles studies, handles senders that do not signal completion cleanly, applies backpressure, passes compressed transfer syntaxes through, and spools safely when the link drops.

• Cloud ingest. A DICOMweb STOW-RS receiver with bounded-memory multipart parsing, study-UID cross-checking, and exactly-once downstream firing per request.

• Tenancy and privacy at the boundary. Customer resolution with no silent default, DICOM PS3.15 confidentiality rules, fail-closed checks before data leaves the customer network, and a re-identification store under customer-held keys.

• The outbound lanes. JSON callbacks, dictation-system autotext, and HL7 v2 ORU^R01 over MLLP, shaped to the interface engine the customer actually runs.

• Visibility and correctness. Per-study tracing, three-component integration health, configuration-drift checks, and CI for both gateway repositories in week one.

• The technical truth in partner review. Precise answers on transport, retention, incident handling, and what the system does and does not do.

The problems are the point. One recent task peaked at roughly nine times the request-body size and was OOM-killed before application logging could catch it. Another integration applied an allowlist to the wrong egress address and failed halfway through a transfer. Both looked like something else. If outside-in debugging, packet captures, request tracing, and reading protocol specifications sound like a good week, this is your kind of role.

Who thrives here:

• You have implemented a protocol from its specification, not only called a library that wrapped one.

• You have run software inside infrastructure you could not SSH into.

• A bare 502 with no application log makes you widen the system boundary, not add another log line and hope.

• Fail-closed is your instinct when patient data is involved.

• You can write a document a hospital IT reviewer will trust because it is precise, traceable, and honest.

Helpful, not required: DICOM, PACS, VNA, or teleradiology systems; HL7 interface engines; pynetdicom, pydicom, Orthanc, dcm4che, or DCMTK; networking, embedded, or payments systems; AWS and Terraform; healthcare or another regulated environment; de-identification or privacy engineering; confidential computing or hardware attestation.

Protocol literacy, outside-in debugging, and fail-closed instincts. That's the bar. Background is a multiplier.

Stack: Python, FastAPI, pytest · pynetdicom, pydicom · DICOM PS3.15 and PS3.18 · DICOMweb STOW-RS · HL7 v2 over MLLP · AWS, Terraform, Docker · shell and installers that work on locked-down hospital systems

Location: Boston (hybrid), US-remote, or remote international, placed to match the account base. This is a customer-facing role: you join hospital IT and clinical reviews, stay reachable during that customer's clinical day, and travel occasionally for an on-premise install, so where you sit follows where the customers are. Your location is a disclosable sub-processor location in our customer agreements.

Compensation (US): $140,000 to $180,000 base plus approximately 10% discretionary bonus. Equity may be offered to top candidates.

Compensation (international, remote): cash-weighted base, tiered by the country the work is performed in. Tier A $110,000 to $150,000 · Tier B $80,000 to $120,000 · Tier C $68,000 to $100,000. No equity on international offers. No visa sponsorship needed, none implied.

Benefits (US): health, dental, vision, and 401(k). International is contractor or employee via an employer of record, depending on your country.

The team: researchers and engineers who have shipped FDA-cleared production systems, working directly with fellowship-trained radiologists. Small enough that the integration path has your name on it, rigorous enough that the evidence has to hold up.

To apply: Use the application form below. Alongside your CV and GitHub, you'll answer one short question: the last integration bug you chased that turned out to be nothing like what the error message said.