Вакансия из Hirify Global, списка международных tech-компаний Для мэтча и отклика нужен Plus
Пожаловаться
86
Крутая вакансия
развернуть
Чётко описанная исследовательская роль даёт возможность влиять на реальные продукты и развиваться в AI, но формат стартапа и ответственность за весь цикл могут означать неопределённость и высокую нагрузку.
?Ответственность за весь цикл
Современный ML-стек
Прикладной AI
?Риски стартапа
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Подключить Plus
Описание вакансии
Текст:
Hirify AI/Оригинал
TL;DR
Applied Researcher (Computer Vision/Video AI): Developing and evaluating production-ready computer vision models for industrial video analytics with an accent on data curation, reproducible experimentation, and accuracy, robustness, and latency optimization. Focus on diagnosing model failures, building rigorous evaluation suites, and converting research prototypes into monitored production systems.
Location: Zurich, Switzerland; hybrid
Company
hirify.global develops video AI for manufacturers, connecting to factory cameras to detect production deviations, quality issues, and safety risks and trigger alerts or automated actions.
What you will do
Set and execute the research agenda for a video-AI platform, from problem framing and data curation through prototyping and evaluation.
Design reproducible experiments, maintain configurations, seeds, artifacts, and documented research workflows.
Diagnose failures across data, augmentations, labels, and training dynamics, then validate targeted fixes.
Define datasets and error sets, manage sampling and imbalance, and partner on labeling and quality assurance.
Build evaluation suites and dashboards covering accuracy, robustness, latency, cost, trends, and failure clusters.
Turn promising research into production-ready models with Engineering, including testing, packaging, canary releases, monitoring, and cost and latency optimization.
Requirements
2+ years of experience working with real-world image or video models.
Master’s or PhD in Computer Science or a related field, with strong computer vision or machine learning knowledge.
Strong Python and PyTorch skills, including clean training and evaluation loops.
Hands-on experience with ML experiment platforms such as ClearML.
Practical knowledge of augmentations, class imbalance, sampling strategies, label quality, and active-learning basics.
Fluency with mAP, F1, IoU, ROC-PR, calibration, robustness testing, and dataset/version management for reproducible comparisons.
Ability to communicate findings as clear recommendations and collaborate with Engineering, Product, and GTM.
Nice to have
Kaggle experience.
Temporal models such as 2D/3D CNNs or transformers, multi-object tracking, and handling motion or occlusion.
Pretraining, VLM adapters/LoRA, or lightweight fine-tuning of foundation models.
Distributed training with DDP/FSDP, HPO frameworks such as Optuna or Ray Tune, and compute budgeting.
Distillation, pruning, quantization, ONNX Runtime, TensorRT, and edge or real-time deployment.
Labeling operations, weak supervision, synthetic data generation, and augmentation pipelines.
Video analytics experience in industrial or safety-critical environments.
Culture & Benefits
Ownership of vision systems running continuously in real factories and helping reduce downtime and waste.
Growth across applied computer vision, production software, and MLOps with support from an experienced technical team.
Flat hierarchies, a get-things-done mentality, and collaboration with colleagues from around the world.
Competitive salary, meaningful stock options, and comprehensive benefits.
Modern office in central Zurich.
Будьте осторожны
Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →