AI Engineer (Computer Vision & Edge) — Mid/Senior

About AnalitixAnalitix turns the security cameras a business already has into people analytics: how many come in and go out, how long they wait, and where they linger. It never records or stores video. A small NVIDIA Jetson device on site runs real-time detection and tracking, and sends only anonymous events to the cloud. Clients see the results on a dashboard. Our clients are in retail, hospitality, and passenger transport.The roleYou will own the vision models from start to finish: from raw footage to a TensorRT engine running reliably on a Jetson at a client site. You will work closely with our software engineers, and you will have real influence over how we build and measure models.What you’ll doTrain, fine-tune and evaluate object detection models (RF-DETR, YOLO) for people and new use casesBuild and improve the data pipeline: draft labels with foundation models (e.g. SAM 3), review them in CVAT, and curate datasetsDesign evaluations that reflect real outcomes, not just mAP. For example, measure per video and count false alarms on footage without the targetExport models to ONNX and compile TensorRT FP16 engines for Jetson OrinIntegrate and tune models in our DeepStream / GStreamer pipeline (inference, tracking, line and zone analytics)Profile and optimise throughput and latency on edge hardware with many camera streamsWrite down your findings as reproducible measurements, so the team makes decisions on dataWhat we’re looking for3+ years of hands-on computer vision / deep learning, with models you shipped to productionStrong Python and PyTorch; solid grasp of detection architectures, training dynamics and evaluationExperience with model optimisation and deployment (ONNX, TensorRT, quantisation / FP16)Comfort working with Linux, Docker and GPUsYou measure before you conclude, and you can explain a result in plain languageNice to haveNVIDIA DeepStream, GStreamer or Jetson / JetPack experienceMulti-object tracking, line-crossing or zone analyticsLabelling workflows (CVAT, auto-labelling with foundation models)Experience with privacy-by-design systems (GDPR)What we offerProduction-grade computer vision and edge AI systems: your models run on real cameras at real client sites, not just in a notebook€2,800 – €5,000 gross per month, depending on experienceA small team where you shape the ML roadmapModern stack: RTX 5090 workstation for training, Jetson Orin in the field, DeepStream 9 / TensorRT 10Office in Split, with remote work possibleHow to applyApply directly via LinkedIn with your CV or LinkedIn profile.