🚀 IntelliProve builds a software platform that turns a face video into vital signs. The work ahead is to make the backend and cloud infrastructure behind that platform genuinely production-grade. Reliable enough to carry medical traffic, traceable enough to pass an audit, and ready to scale with the product. This role owns the Python/FastAPI backend and the AWS infrastructure that make that possible.

Working at IntelliProve

At IntelliProve we're making proactive and preventive health monitoring effortless, turning a short face video into objective vital-signs data for both consumer apps and clinical tools.

We're a 14-headed team in Ghent building two products on the same rPPG technology: a consumer wellbeing app that's live today, and a medical device heading for CE marking. Working on both in parallel keeps us close to real users and real regulators at the same time. Feedback loops are short, and decisions get made by the people doing the work.

We value ownership, curiosity and clarity. People at IntelliProve take responsibility for what they ship from design through deployment, ask questions instead of guessing, and prefer clear, maintainable solutions over clever ones. You'll work side by side with engineers, ML researchers and regulatory colleagues, and your input will shape how the platform evolves.

The role at a glance

• Full-time role, open to a freelance arrangement
• Dedicated computer vision ownership of the rPPG signal-extraction pipeline
• Close collaboration with our CTO, ML colleagues, backend engineers, QA lead and regulatory team
• Direct impact on the rPPG engine that powers IntelliProve's vital-signs measurement
• Ghent office, up to 2 days remote per week

What you'd be working on

rPPG signal extraction

• Improve how the pipeline detects and tracks the facial regions that carry the strongest blood-volume-pulse signal
• Push the robustness of signal extraction against the noise and variability of real-world consumer-camera data
• Tackle motion compensation and ROI-tracking problems that rarely have clean off-the-shelf answers

Models & architectures

• Design and train deep-learning CV models in PyTorch, with a strong emphasis on temporal architectures (3D CNNs, temporal convolutions, transformer-based temporal models), since the blood-volume-pulse signal lives in the temporal dimension
• Find ways to leverage the unlabelled video data captured from live users to make models more robust over time

Production & medical validation

• Bring models into production: real-time inference, model compression, ONNX, sensible deployment on our AWS stack
• Improve our MLOps so model training, evaluation and rollout scale with the product
• Produce the documentation, validation evidence and explainability artefacts the medical pipeline requires for CE marking

Who are we looking for

• Several years of professional experience in computer vision and/or machine learning at a medior level (+3 years), or equivalent practical experience with a relevant educational background.
• Strong computer vision background. Hands-on experience with classical CV (OpenCV, colour spaces, signal/frequency analysis, ROI tracking) and modern deep-learning CV (CNNs, vision transformers). Comfortable with face detection, landmarking, segmentation and tracking
• Real experience with video and temporal modelling. You've extracted information from frames over time using 3D CNNs, temporal convolutions, RNN/LSTM/Transformer-based temporal models, or optical flow. Bonus if you've worked on physiological signal extraction, motion compensation, or any problem where the temporal dimension carries the information rather than the spatial one
• PyTorch as your primary framework. Equivalent fluency in TensorFlow or JAX is acceptable if you're happy to move to PyTorch
• Comfortable with image and video data pipelines and the practical realities of working with consumer-camera footage
• Solid engineering hygiene and a research mindset; comfortable working iteratively when the path to a solution isn't laid out for you

Nice to have

• Direct experience with rPPG or physiological signal processing: POS, ICA-based methods, or recent end-to-end rPPG architectures (PhysNet, DeepPhys, RhythmNet, Contrast-Phys)
• Real-time inference experience: model compression, quantisation, ONNX
• MLOps on a major cloud (we use AWS). Our setup is intentionally basic today and needs to scale with the product
• Explainable AI techniques. Useful for documenting ML components for medical certification: the clearer the explanation of how a model works, the smoother the regulatory path

A few honest self-checks

• You're comfortable with research ambiguity, willing to read papers, prototype, fail and iterate when the path to a solution isn't obvious
• You know where rigour matters (medical paths, validation evidence) and where pragmatism wins; you're systematic without being slow
• You take operational ownership of what you ship. Models drift, edge cases surface, and you don't throw them over the wall
• You communicate clearly with non-ML colleagues (engineering, regulatory, product) and prefer precise questions over vague ones
• You're comfortable to say "I don't know" and go find out

What we offer

• Full-time role (employee or freelance) in a healthtech company building toward CE marking
• Dedicated ownership of the computer-vision layer in the rPPG engine
• Competitive salary and additional benefits, aligned with your experience, plus stock options (ESOPs) and KPI-tied bonuses via Payflip
• Small team, short feedback loops, high autonomy
• Flexible working setup. Up to 2 days remote per week, with the rest at our Wintercircus office in Ghent, a vibrant working hub with lunch & learns, masterclasses, campus open hours and an on-site gym
• Software that ends up in clinical settings, not yet another SaaS dashboard

Practical details
Working language is English; Dutch is preferred. The role is based at our Ghent office and requires the right to work in Belgium.

Interested? Drop us your resume and a short note on what caught your eye at [email hidden].