Material Bank is seeking a Staff Applied Scientist to drive building search and recommendations from the ground up. We are the largest marketplace for architecture and design professionals, and we are building a discovery platform that helps them find materials across hundreds of brands and surfaces intelligent recommendations as they explore. Search and recommendations are how that discovery happens. The surface is wide open, the right techniques for our domain are still unsettled, and the person in this role is the one who figures them out and builds them.

This is a hands-on research role. You will not hand formulas to an engineering team and walk away. You will explore the data, form the hypothesis, prototype the approach, and prove it works, then partner with engineering to take it to production.

What you'll do

• Improve search retrieval and ranking. Design, tune, and evaluate how we retrieve and rank search results so members find what they need and discover what they did not know to ask for. Bring real depth on ranking.
• Build our recommendation systems from the ground up. Intelligent recommendations are central to the discovery experience we are building. Design robust recommendation algorithms, model member and project intent, and figure out what to surface, when, and where across the experience.
• Explore the data to find the strategy. Run exploratory analysis on behavioral and catalog data to figure out which approaches fit our members and our catalog. The answers are not known yet; a large part of this job is asking the right questions and finding them in the data.
• Apply advanced query understanding. Improve query parsing, intent recognition, and semantic understanding for diverse, messy real-world queries, including the long tail.
• Raise data quality. Search and recommendation quality is downstream of catalog quality. Identify the data problems that hold relevance back and help fix them.
• Do R&D you can ship. Invent the formulas, models, and re-ranking approaches and build working versions yourself. Iterate and reformulate quickly, then collaborate with engineering to productionize what proves out.
• Iterate visual and color approaches. Evaluate image-based search and color matching, including perceptual color, and decide what is right for us.
• Measure what matters. Build and extend evaluation frameworks and success metrics so we know whether a change actually helped, and design experiments as traffic grows.
• Collaborate across functions. Work closely with product, engineering, and catalog to turn research into shipped improvements.

What you'll need

• A track record building search, recommendations, or ranking systems that shipped to real users and moved the numbers. You have done this work before, not just studied it.
• An advanced degree in a quantitative field such as computer science, machine learning, statistics, or applied math, or equivalent hands-on experience that got you to the same place.
• Deep understanding of search and recommendation systems, especially retrieval, ranking, and relevance. This is the core of the role.
• Strong query understanding and NLP, including semantic and embedding-based retrieval.
• Real exploratory and experimental rigor: statistical modeling, experimental design, and hypothesis testing, plus comfort working in ambiguity where the right approach is not yet known.
• Hands-on building ability. You prototype and implement your own ideas. We do not care which language you are strongest in.
• Some experience deploying ML or data systems to production, or partnering closely on it.

Nice to have

• Experience with multimodal or image-based search, and with color or perceptual science.
• E-commerce or marketplace search and recommendations experience.
• Experience managing vendor and home-grown systems in conjunction with one another.

Why this role is unusual and exciting

Most search science roles plug into a mature stack with the strategy already decided. This one does not. You set the approach, build it, and prove it, with a clear runway toward owning more of the search and recommendations system over time. If you want the surface area of a founding scientist for a search function inside an established, well-funded company, this is that.