About the role
Job description: Adobe Fonts helps millions of creatives find and use the right type. This role owns how users discover fonts — search, relevance, recommendations, and ranking — and ships it end to end. You are a software engineer first, you turn what product and design want into reliable, well-crafted production experiences, own them in production (including on-call), and apply machine learning — embeddings, ranking, relevance, fine-tuning — as a tool for delivery, not a research agenda.
What you'll do - Build, ship, and iterate on font-discovery experiences with product and design — from spec to production — with strong software craft: clean design, testing, code quality, and code review. - Own search relevance and ranking: candidate generation, retrieval (embeddings / vector search), learning-to-rank, re-ranking, and relevance / search optimization. - Build and improve recommendation and personalization surfaces (related fonts, pairings, contextual suggestions). - Own what you ship in production: latency, scale, reliability, observability, and support for the search and recommendation services. - Design the systems around the models: feature pipelines, low-latency serving, evaluation harnesses, A/B tests, and closed feedback loops from impressions, clicks, and usage. - Define and defend the metrics — offline relevance and online outcomes — and debug the gap between them.
What we're looking for Software engineering (your core): - Strong software fundamentals and best practices — design, testing, code quality, code review — plus system design and production maturity (latency, scale, reliability, observability, on-call). - Ships and owns services end-to-end; pragmatic about build-vs-buy and tradeoffs. Applied ML depth in search / relevance / recommendations: - Hands-on, working understanding of embeddings, models and fine-tuning, retrieval and ranking (ANN / vector search, learning-to-rank), relevance and search optimization, and recommendation systems. - Solid evaluation instincts — metrics design, error analysis, offline↔online debugging — and comfort closing the loop with behavioral data (clicks, impressions, usage). - Applies ML judgment to ship better discovery; depth without a research mandate. - Delivery focus and product sense, end-to-end ownership, systems thinking, and cross-functional execution with product and design.
