Ai engineer in computer vision

UFORCE exists to make aggression unaffordable. Founded in Ukraine and headquartered in London, UFORCE is a defence-technology company with operations across Europe, the United States and Asia. UFORCE builds uncrewed vessels, ground robotics and aircraft, autonomy software, and the command-and-control layer that operates them together. Each platform is deployable on its own, and stronger as part of the system. Its product lines include the MAGURA family of uncrewed surface vessels, the NEMESIS family of strike platforms, the LIUT family of ground robotic platforms, and counter-UAS systems.

• Own the ML adaptation pipeline from raw customer data to trained, evaluated, deployable models.
• Build foundation-model-assisted labeling workflows using tools such as SAM-2, Grounding DINO, open-vocabulary models, LLM steering, and human review.
• Design self-serve operator workflows using yes/no/maybe feedback and natural-language corrections.
• Create versioned datasets with lineage, data cards, label provenance, class distributions, and known gaps.
• Build QA methods that catch systematic pseudo-label errors, missing annotations, and long-tail data gaps.
• Develop active-learning loops that prioritize the highest-value frames for limited operator review.
• Build reproducible train/eval pipelines with experiment tracking, model packaging, and promotion gates.
• Design evaluation around held-out anchor sets, leakage prevention, baselines, slice metrics, and automated approve/reject decisions.
• Package the toolchain for on-prem, air-gapped, regulated, or customer-held environments.
• Close the field-failure loop by feeding live failures back into the next tune cycle.
• Lead and grow a small specialist team around curation, auto-labeling, deployment, and onboarding.

• 5+ years building production ML, AI, or computer-vision systems.
• Strong Python and PyTorch.
• Experience owning ML data pipelines, training pipelines, or evaluation infrastructure.
• Deep CV data experience: detection, segmentation, annotation taxonomies, dataset curation, and data quality.
• Hands-on experience with model-in-the-loop or foundation-model-assisted labeling.
• Familiarity with tools such as SAM-2, Grounding DINO, FiftyOne, and annotation platforms.
• Strong understanding of evaluation: held-out sets, leakage prevention, baselines, slice metrics, and promotion gates.
• Experience with QA sampling, label-noise analysis, missing annotations, IAA, or alternatives when only one operator is available.
• Experience with active learning or other methods for prioritizing labeling effort.
• Ability to reason about dataset economics: quality vs. quantity, long-tail coverage, and cost-per-useful-example.
• Experience with dataset versioning, lineage, experiment tracking, model registries, or data cards.
• Strong ownership, communication, and systems thinking.
• Experience leading engineers as a tech lead, staff engineer, or small-team manager.
• Будет плюсом: Synthetic data, sim2real, or domain randomization experience; EO / IR / LWIR, remote sensing, maritime imagery, or small-object detection experience; Privacy-preserving ML, federated learning, on-prem, or air-gapped deployment experience; Experience building self-serve ML platforms or tools for non-expert users; Experience with lakeFS, DVC, MLflow, Weights & Biases, Kubernetes, Kubeflow, Flyte, Dagster, Airflow, or Argo; LLM application, context engineering, structured output, or LLM evaluation experience; Exposure to radar, AIS, EO/IR fusion, tracking, sensor fusion, robotics, autonomy, UxV, defence tech, C2/C4ISR, or tactical systems

• Условий в вакансии нет