• Train, fine-tune, and adapt CV and vision-language models (detection, classification, VLMs) to document and image tasks
• Work with datasets end to end: define data requirements, prepare and structure data, support labeling workflows, and improve data quality
• Optimize models for production — accuracy, latency, cost, and robustness to real-world image quality (noise, skew, lighting, low-res scans and phone photos)
• Collaborate with engineers to bring trained or adapted models into real production use cases

What makes you a great fit:

• Demonstrated experience in deep learning and transformers models
• Hands-on experience with OCR and document understanding (text detection and recognition, layout analysis, information extraction) using VLMs and classical CV models
• Hands-on experience training, fine-tuning, or adapting CV models for practical use cases
• Proficiency in frameworks like PyTorch, with CV tooling such as OpenCV / torchvision
• Understanding of how model quality depends on dataset design, labeling quality, and data preparation
• Ability to go deeper than standard training notebooks and make informed decisions during training and optimization
• Strong foundation in data structures, algorithms, and software engineering principles
• Familiarity with LLM training and fine-tuning methods, such as distillation, supervised fine-tuning, and policy optimization
• Experience with content understanding or abuse/fraud prevention is a plus
• B1 or higher English level for effective communication with an international team