> Markdown version of [/jobs/ext/3178540-ai-engineer](https://www.wearedevelopers.com/jobs/ext/3178540-ai-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # AI Engineer - **Company:** Sitemark - **Location:** Leuven, Belgium - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** Application Programming Interfaces (APIs), Artificial Intelligence, Computer Vision, Backup Devices, Python (Programming Language), PostgreSQL, Object Detection, TypeScript, Management of Software Versions, ReactJS, Delivery Pipeline, Deep Learning, Machine Learning Operations - **Published:** September 25, 2026 - **Apply:** https://www.indeed.com/viewjob?jk=dc9807f33fcee936 ## About the Role * Strong applied computer vision / deep learning experience - you've trained, fine-tuned, and debugged CV models, not just called APIs, and you understand what's happening inside them. * Hands-on with the experimental loop: dataset curation, augmentation, training, error analysis, iteration. When results are bad, you know how to diagnose why. * A pragmatic, product-oriented mindset - you reason about how a model will actually be used, what "good enough" means for the business, and the shortest path to a real result. * Strong fundamentals and clean engineering instincts. You write code meant to live in production - readable, testable, maintainable - not just notebook scratch. * Motivated to grow into the integration and MLOps side, and comfortable touching code beyond the model itself. (You don't need to be a senior full-stack engineer on day one.) * High intelligence and learning velocity - we care more about how you think and how fast you grow than years on a CV. * Comfortable working in English in a small, fast-moving team. ## Description At Sitemark, AI sits at the heart of the product. Drones and robots capture vast amounts of imagery from renewable energy sites, and our computer-vision models turn it into the answers customers act on. We need someone to scale that capability so it ships reliably and moves real business metrics. As our AI/ML Engineer, you own the AI/ML side of the platform: training and improving the computer-vision models behind our products, and making sure they actually ship and perform in production. You'll raise our throughput across model implementation, training runs, and dataset iteration - directly unblocking the team and our customers. We're looking for a pragmatic engineer-scientist, not a paper-chaser. Models exist to solve real problems: if an off-the-shelf model fine-tuned on our data does the job, that's a great answer. We care about results in the product, not novelty in a paper. No solar or energy background required - we'll teach you the domain; curiosity matters more. You'll report to the Head of Product & Engineering, with coaching and technical sparring from the Engineering Lead, and work in cross-functional squads alongside platform engineers, the product team, and - closely - our operations teams and customers. What you'll do: * Train, fine-tune, and ship computer-vision models for tasks like thermal anomaly detection and classification, defect detection on high-resolution imagery, object detection on drone imagery, and stitching/co-registration support. * Level up the MLOps backbone that lets us ship reliably: experiment tracking, reproducible training, dataset versioning, a model registry, deployment pipelines, production monitoring, and a feedback loop from labeled operations data back into training. This is where AI meets engineering, and it's a big part of what makes the role impactful. * Run the full experimental loop end to end: curate and improve datasets, design training runs, analyse errors, and iterate. * Take on the harder architectural problems when they matter - for example, models that reason over large spatial context (an entire site, not just a tile) where a standard fixed-resolution detector falls short. * Integrate models into the product end to end. A model isn't done when the metric looks good - it's done when it's running on real data in the platform and making the team or the customer faster. * Choose problems and approaches based on business impact - what actually moves the needle for our products and operations. What success looks like after 6-12 months: * Your models are running in production on real data, moving the metrics our customers and operations teams care about. * The MLOps backbone is solid - training is reproducible, experiments are tracked, and shipping a new model is routine, not a fire drill. * A feedback loop runs from labeled operations data back into training, so the models keep getting better on their own cadence. * Throughput across model work, training runs, and dataset iteration is meaningfully higher than when you started. * AI is unblocking the roadmap, not bottlenecking it - the gaps we hired you to close are closed., * Detection, segmentation, keypoint, or multi-scale architectures applied to large or high-resolution images. * Production MLOps: experiment tracking, reproducible training, model registries, monitoring. * Full-stack experience (Python, TypeScript, React, Postgres) - you'll get plenty of chances to use it. * Weakly- or self-supervised learning, active-learning loops.