описание
Mad Devs works on complex real-world engineering challenges and applied AI projects involving engineering data, production constraints, and customer requirements.
задачи
- Design and develop computer vision and machine learning solutions for architectural drawings, technical documents, CAD, and BIM data
- Build pipelines combining raster images, vector geometry, structured engineering data, OCR, and ML models
- Reconstruct structured objects from geometric primitives, including topology and spatial relationships
- Analyze dense technical drawings and documents based on their structure and semantics, not only visual object detection
- Design and maintain evaluation pipelines with ground-truth data, quality metrics, regression testing, and validation on unseen data
- Build human-in-the-loop workflows with calibrated confidence scores and escalation of uncertain cases for human review
- Evaluate modern approaches to floor-plan and technical-document understanding, including detection, segmentation, transformer-based models, and raster-to-vector reconstruction
- Develop production-ready inference services with CPU inference by default and optional GPU acceleration
- Make training and inference reproducible, including model versioning, packaging, and deployment requirements
- Transfer models to customers for independent deployment and further training
- Provide technical leadership on ML and computer vision decisions within a small engineering team
- Communicate directly with technical stakeholders and explain ML limitations, trade-offs, and decisions in English
- Challenge proposed approaches when needed and recommend better alternatives based on technical evidence
требования
- 5+ Years of professional experience in machine learning or computer vision, including taking ML solutions into production
- Strong knowledge of computer vision, machine learning, and model evaluation
- Proven experience validating model quality using measurable metrics and real-world data
- Practical experience with vector geometry, structured data, topology, and spatial relationships
- Experience processing technical documents, diagrams, or drawings where structure and relationships matter
- Experience designing evaluation pipelines, including ground-truth datasets, regression testing, and generalization testing
- Experience building human-in-the-loop ML systems and handling uncertain predictions
- Experience making ML training reproducible and preparing models for independent deployment or further training
- Understanding of model and dataset versioning and relevant licensing considerations
- Experience providing technical leadership within a small engineering team
- Experience communicating directly with technical customers or stakeholders
- English at B2–C1 level with strong spoken communication skills
- Russian at C1–C2 level or native proficiency
- Будет плюсом: experience with architectural or construction drawings, BIM models, quantity takeoff, or cost estimation; familiarity with research in floor-plan recognition, raster-to-vector reconstruction, or structured document understanding; experience parsing CAD and BIM formats, including malformed, incomplete, or inconsistent exports; experience with OCR for European languages and industry-specific terminology; experience optimizing ML inference for CPU-only environments; experience deploying ML services in isolated customer environments without external dependencies; experience with graph-based representations of engineering structures or building layouts; ability to read technical drawings in German or French; German or French at B1–B2 level
условия
- Flexible working hours
- Long-term projects
- Salary in dollars
- Professional communities
- Onsite business trips
- Training budget
- Paid conferences
навыки
компания
Международная IT-компания Mad Devs, основанная в 2016 г., разрабатывает кастомное ПО: веб- и мобильные приложения, frontend и backend, DevOps и QA, блокчейн-решения и продукты машинного обучения, обеспечивая прозрачность процессов и тесное взаимодействие с клиентами
Подробнее о компании →