Visual AI Engineer

About Master Art Index Master Art Index (MAI) is building the definitive intelligence layer for the global art market. We develop AI-driven valuation models, structured databases, and financial-grade data infrastructure for blue-chip modern and contemporary artworks. Our platform sits at the intersection of finance, technology, and art — enabling institutional-quality analysis in one of the world’s most opaque asset classes.

We are a cross-functional team of AI engineers, computer vision specialists, product managers, and full-stack developers. We move fast, value ownership, and build things that don’t yet exist. Role Overview As Visual AI Engineer, you will own two of MAI’s most strategically critical AI capabilities: – Artwork Recognition — building and continuously improving the pipeline that identifies an artwork from a user-submitted photograph and matches it accurately to the corresponding record in MAI’s database. – Visual Attribute Extraction — developing the systems that decompose each artwork in our database into structured, labeled visual components — medium, style, composition, subject matter, colour palette, and more — enabling downstream valuation modeling and peer artwork identification.

Both capabilities sit at the core of MAI’s product and data strategy. You will not be maintaining existing infrastructure — you will be advancing state-of-the-art systems in a domain where labeled data is scarce and precision requirements are high. You will work closely with the Head of Art Research, AI Engineers, the Art Market Data Engineer, the Product Manager, and Full-Stack Developers, reporting to the Head of AI Engineering.

Core Responsibilities Artwork Recognition – Design, build, and iterate on the end-to-end artwork recognition pipeline — from image intake and preprocessing through feature extraction, similarity search, and database record matching. – Develop and maintain image embedding models and vector search infrastructure (e.g. CLIP, FAISS, or equivalent) to support accurate and scalable visual matching. – Handle the edge cases: partial images, poor lighting, varied angles, and reproduction photos — building robustness into every layer of the pipeline. – Integrate external lookup pipelines and API-based metadata retrieval for cases where a local match is not found or confidence is insufficient. Visual Attribute Extraction – Build and refine pipelines that extract structured visual attributes from artwork images — including medium, artistic style, compositional structure, subject matter, dominant palette, and surface characteristics. – Work alongside the Head of Art Research to define and validate the attribute taxonomy, ensuring extracted data is art-historically grounded, consistent, and usable for valuation modeling. – Leverage multimodal and vision-language models (e.g.

CLIP, LLaVA, or equivalent) to generate semantic descriptions and structured tags at scale. – Design annotation schemas and contribute to training data curation strategies for domain-specific fine-tuning. Engineering & Infrastructure – Package CV and ML models as production-grade microservices (FastAPI, Docker) and deploy on AWS (ECS, SageMaker, Lambda, S3, or equivalent). – Write clean, modular, testable Python code built for handoff and long-term maintainability. – Contribute to MLOps practices — model versioning, evaluation pipelines, and retraining workflows. Core Requirements – 3–6 years of hands-on experience with computer vision pipelines — object detection, image segmentation, preprocessing, and feature extraction (e.g.

YOLO, Detectron2, OpenCV). – Proven experience with image embedding and similarity search, including approximate nearest-neighbour search at scale (e.g. CLIP, FAISS, or equivalent vector search approaches). – Strong proficiency in PyTorch (preferred) or TensorFlow; solid working knowledge of CNNs and Vision Transformers (ViT). – Experience applying multimodal or vision-language models (e.g. CLIP, LLaVA, or similar) to extract semantic attributes and structured information from images. – Ability to design and implement metadata enrichment pipelines — querying external APIs or structured web sources when local image matching yields insufficient results. – Solid software engineering fundamentals — clean, production-ready Python with attention to modularity, testability, and handoff to downstream systems. – Experience packaging ML models into microservices (FastAPI preferred) and containerising with Docker. – Hands-on experience deploying ML workloads on AWS (ECS, SageMaker, Lambda, S3, or similar); comfort with containerised and infrastructure-as-code deployment patterns. – Experience setting up and maintaining model training pipelines from scratch — covering data strategy and augmentation, loss function design, training infrastructure, evaluation frameworks, and iterative improvement cycles.

Good to Have – Familiarity with vector databases for similarity search at scale (e.g. Pinecone, Weaviate, pgvector). – Experience with OCR pipelines (e.g. Tesseract, AWS Textract) for text extraction from artwork images — signatures, labels, plaques. – Experience fine-tuning foundation vision models on small or noisy domain-specific datasets. – Domain familiarity with art history, artistic movements, or iconographic conventions — useful for training data curation, annotation schema design, and feature relevance judgments. – Familiarity with MLOps practices — model versioning, monitoring, and retraining pipelines (e.g.

MLflow, Weights & Biases). – Experience with agentic or RAG-based approaches for metadata enrichment from online sources. – Background in medical imaging or other high-stakes, data-scarce visual domains (e.g. pathology, radiology, satellite imagery) — the precision culture, domain-expert collaboration model, and constrained annotation environments transfer directly to MAI’s context. What We Offer – A technically differentiated problem set: artwork recognition and semantic visual analysis in a data-scarce domain with high precision requirements. – Direct collaboration with a Head of Art Research — a rare cross-disciplinary working relationship between ML engineering and deep domain expertise. – Ownership of two strategically critical AI capabilities from architecture through production. – A lean, cross-functional team where technical decisions carry real weight and ship quickly. – Competitive base compensation: $90,000 – $100,000 CAD, commensurate with experience. – Modern offices in the vibrant Old Port of Montreal – On-site (4 days) / Remote (1 day) Team & Reporting This role reports to the Head of AI Engineering and works in close collaboration with: – Head of Art Research – AI Engineers – Art Market Data Engineer – Product Manager – Full-Stack Developers