SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
FYUL (formed in 2024 from the merger of Printful, Printify, and Snow Commerce) is seeking a Staff Data Scientist to lead the technical direction and scaling of core AI capabilities across two interconnected domains: Search, Ranking & Recommendations, and Generative AI & Visual ML.
In this role, you will lead the architectural direction for candidate retrieval, multi-objective ranking, and learning-to-rank (LTR) models across raw product and design recommendations. You'll advance personalization engines incorporating sequence embeddings of merchant history, niche trends, and sales performance signals. You'll connect downstream merchant sales and conversion data back into upstream design generation and candidate ranking models to continuously improve recommendation relevancy.
You'll oversee image generation pipelines combining LLMs and multimodal models with deep learning architectures (diffusion models, UNets, Vision Transformers) to generate high-converting designs for merchants. You'll deepen classical computer vision capabilities (OpenCV) alongside deep learning vision models (ResNet, CNNs, YOLO/segmentation) for delayering, automated background separation, print-area quality analysis, and asset validation.
As a product engineer with full technical autonomy across the stack, you'll handle foundational model selection (fine-tuning, prompt engineering, self-hosting) through serving pipelines and frontend workflows. You'll rapidly prototype, iterate, and ship complete agentic pipelines and ML features using AI-assisted development tools.
You'll define and refine online/offline evaluation frameworks (NDCG, MRR, visual quality metrics, merchant conversion uplift) and oversee continuous A/B testing strategy. You'll guide strategic make-vs-buy decisions, API vs. self-hosted fine-tuning trade-offs, and inference optimization. You'll set standards for applied ML across the organization, mentor senior individual contributors, run design reviews, and foster a culture of shipping measurable business impact.
The tech stack includes PyTorch, OpenCV, Hugging Face Transformers, Diffusers, LangChain, LangGraph, vector databases (Qdrant/Pinecone/FAISS), Flyte for workflow orchestration, CometML for experiment tracking, and Snowflake for data warehousing.
QUALIFICATIONS:
- 8+ years of hands-on experience building, deploying, and scaling production ML systems with proven staff-level technical leadership
- Deep expertise in modern search, candidate generation, and recommendation architectures (Two-Tower models, ANN vector indexing, Transformer rerankers, learning-to-rank algorithms)
- Hands-on experience with classical CV (OpenCV) and modern deep vision architectures (CNNs, Vision Transformers, diffusion models, multimodal LLMs/CLIP) for image processing and generative workflows
- Proven ability to build closed-loop ML systems that feed downstream business outcomes (clicks, conversions, sales) back into upstream model training and generation logic
- Fluency in Python and the modern AI/ML stack, anchored in PyTorch for deep learning and vision models, alongside LLM orchestration and agentic frameworks (LangChain, LangGraph)
- Practical experience with production serving patterns (FastAPI, event-driven async processing), MLOps tools (CometML, Flyte), and cloud data warehouses (Snowflake)
- Comfort leveraging AI tools across the workflow (Gemini for ideation and research, Claude Code and Claude Cowork for rapid prototyping and production code development)
- Proven track record of taking full technical ownership of end-to-end features, treating ML systems as products and making architectural decisions across the whole stack
Nice to have: experience in e-commerce, two-sided marketplaces, or creator platforms.
