Senior Machine Learning Engineer

Role Overview: You will be responsible for owning the full ML stack capable of transforming raw dielines, PDFs, and e-commerce images into a self-learning system that can read, reason about, and design packaging artwork. Your tasks will involve building data-ingestion & annotation pipelines for SVG/PDF to JSON conversion, modifying model heads using technologies such as LayoutLM-v3, CLIP, GNNs, and diffusion LoRAs, training & fine-tuning on GPUs, shipping inference APIs, and collaborating closely with packaging designers and a product manager to establish yourself as the technical authority on deep learning within this domain. Key Responsibilities: - Data & Pre-processing (40%): - Write robust Python scripts for parsing PDF, AI, SVG files, and extracting text, color separations, images, and panel polygons. - Implement tools like Ghostscript, Tesseract, YOLO, and CLIP pipelines. - Automate synthetic-copy generation for ECMA dielines and maintain vocabulary YAMLs & JSON schemas. - Model R&D (40%): - Modify LayoutLM-v3 heads, build panel-encoder pre-train models, add Graph-Transformer & CLIP-retrieval heads, and run experiments to track KPIs such as IoU, panel-F1, color recall. - Conduct hyper-parameter sweeps and ablations. - MLOps & Deployment (20%): - Package training & inference into Docker/SageMaker or GCP Vertex jobs. - Maintain CI/CD, experiment tracking, serve REST/GraphQL endpoints, and implement an active-learning loop for designer corrections. Qualifications Required: - 5+ years of Python experience and 3+ years of deep-learning experience with PyTorch, Hugging Face. - Hands-on experience with Transformer-based vision-language models and object-detection pipelines. - Proficiency in working with PDF/SVG tool-chains, designing custom heads/loss functions, and fine-tuning pre-trained models on limited data. - Strong knowledge of Linux, GPU, graph neural networks, and relational transformers. - Proficient in Git, code review discipline, and writing reproducible experiments. In the first 6 months, you are expected to deliver a data pipeline for converting ECMA dielines and PDFs, a panel-encoder checkpoint, an MVP copy-placement model, and a REST inference service with a designer preview UI.,