Software Engineer (Data & AI) – Healthcare

If you want frontier tooling and tech in your daily work, real ownership of what you build, and a domain with real impact – read on. If you've stopped continuously learning, or you'd rather spend weeks deep inside one narrow problem, this isn't the place.
Key facts

Role: mid-level engineer working end-to-end on data/AI systems for healthcare – from classical ML to LLM-based agents at model level, through handling the data pipelines, cloud, and integrations that make them run in production.
Team: work is fully remote, but most of the team is in Poland, in the Poznań area. We meet in person during team retreats on a roughly quarterly basis.
Who you work with: a small, experienced technical team, with 20+ healthcare data/AI projects delivered, our internal medical advisors, and clients directly.
Contract: B2B or mandate, flexible hours.
Languages: Fluent English and Polish – written and spoken. Most of the team and part of our clients work in Polish, so fluent Polish is a hard requirement. We can make rare exceptions here for exceptional profiles – if you believe that's you, say why in one line in your application.
What we do – and why healthcare
We're datarabbit.ai – data and AI partner for healthcare and life sciences organizations. We design and deliver production-grade data and AI systems for medical devices, hospitals, life sciences, and pharma across the EU and US.
Healthcare AI is one of the toughest technical niches today. Compliance requirements, integrations with "heavy" legacy systems, end users for whom a PoC built on "we'll figure it out later" is unacceptable. Many companies steer clear of this space entirely.
We don't. We build data platforms processing terabytes of medical data for research. Voicebots talking to patients after surgery. Systems automating analysis and filling of medical documentation. AI agents for enterprise healthcare. All production-grade, compliant, secure, end-to-end – and we've been at it since before AI was fashionable.
Some of what that looked like in practice:
• NLP/LLM documentation automation pipelines that helped cut the staff time required by ~60% – from ~1.5 hours to under 30 minutes per patient;
• developing frontier imaging models doing aneurysm detection from brain imaging at human-level sensitivity for a neurology startup;
• a cloud-agnostic MLOps platform that let a genomics research team run 40% more experiments in the same time.

Where you fit – what a day may actually look like
Morning – design. The hospital documentation automation system – that fills forms from clinical notes so that humans don’t have to – needs to cover a new type of handwritten document. You sketch where it fits, decide what the extraction step owns and what stays in the form logic.
Then the one meeting of today: a 30-minute sync with the project team. You walk them through the design, they poke at it, and the physician on the team settles which fields the model may leave empty and which it must never guess.
Midday – implementation. You build the extraction step. The boilerplate comes out of agentic coding tools, supported by our internal skills and patterns – but the judgment calls are ultimately yours.
Late afternoon (you took a couple of hours off in the middle of the day for something personal, which is fine here) – you check your changes after deployment to the dev environment. There was a small formatting issue, which you traced in the logs, fixed in the code, and added a regression test for. If needed, the client could see the new feature the same day.
You end the day with the PR submitted and a short write-up in the project channel – if it's reusable, it goes to the shared component library. You also check up updates from another project team in the Slack channel – looks like they'll have something good to show at this week's knowledge sharing session.

Over time
Ownership grows, and so does what you know. In the first months you own a whole component inside a project and pick up the domain – how hospitals actually work, what compliance means for tech in practice.
Around six months you're the engineer a project lead hands a whole stream to, with a client-facing call or two and the first architecture decisions that are yours. After a year you own a project stream with someone more junior alongside, or a piece of internal IP – and the cloud, MLOps or domain certifications you collect on the way are funded and rewarded. If you're heading toward leading, say so. We set goals together every quarter, and when you're stuck, there's always someone to ask.

What we require
•

3+ years in software and ML / AI engineering

•

Problem-solving mindset, not just task execution. If your approach is "I didn't go into IT to talk with people" – this isn't the place. We dig into the client's business and often ask better questions than they do themselves.
• AI power-user mindset. You actively use LLMs, agents, and frontier tooling to automate your own work. You follow the industry regularly, not on a quarterly basis. AI-first isn't a buzzword for us – it's how we work daily.
• Genuine interest in healthcare. No prior healthcare experience required (we'll help you ramp up the domain) – but you need to be authentically excited about building tech that affects how people are cared for. If your "interest" stops at "AI is cool" – we're probably not the right fit.
• Full-stack engineering mindset – Python/AI is the core, but you're fine working with infra, data pipelines, cloud, and surrounding systems when the project demands.
Nice to have
• Knowledge of healthcare domain: specific systems, standards, and formats (e.g. HL7/FHIR, DICOM, Nextflow, etc.)
• Experience with classical ML, deep learning frameworks, and modern LLM stack
• Experience with data engineering and workflow orchestration (Airflow, Prefect, etc.)
• Strong cloud skills (AWS preferred, Azure/GCP welcome)
• Bachelor's, Master's, or PhD in CS, Biomedical Engineering, ECE, or related field
• A GitHub or portfolio – not required, but if you maintain one, it's a definite plus and we'll go through it

What you get
•

Real impact – production systems used by medical clients, actually affecting patients, public health, and the pace of research.

• Frontier mindset – we adopt new models and tools as soon as they make sense. We don't spend quarters getting them approved, like large organizations do.
• Swiss army knife mode – models/AI is one thing, but we work end-to-end: ML, infra, data, client, user-facing apps. For some this is a killer, for others the best job they've had. If you love spending weeks deep in one narrow problem – this isn't us.
• Direct influence on projects and the company's direction – you won't be another cog in the machine.
• Fully remote with flexible hours – need to pick up your kid, hit a doctor's appointment, work better in the evenings? No problem.
• Dedicated training budget and learning opportunities – weekly sessions where the team shares what it's learned; quarterly goal reviews with certifications funded and rewarded; research topics and publications; hackathons participation; and more;
How the hiring process looks:
CV pre-selection by the engineering team, then:
•

Intro call – 5-10 min: background, some additional details about the company and role, your questions.
• A short take-home coding task – a simple one, up to an hour or so max to complete – but you will have a week to comfortably fit it into your schedule.
• A technical interview with two of our engineers.
• A conversation with the CEO.

After the pre-selection, the rest of the process usually takes up to ~2 weeks