Designing Production AI Retrieval Systems
I work with engineering teams on retrieval, ranking, RAG, recommendation, and personalization systems that must meet real-world requirements for relevance, latency, freshness, scale, and cost.
My work focuses on hybrid retrieval, multi-stage ranking, real-time updates, query-time features, tensor-based ranking, and unified serving with Vespa.ai.
BM25 + ANN · Multi-vector retrieval · Rank profiles · Multi-phase ranking · ML ranking · Streaming search
Systems I Design and Evaluate
Search, RAG, Recommendations, and Personalization Powered by Vespa AI Search Platform
Retrieval and RAG
Design document and chunk schemas; combine lexical retrieval, nearest-neighbor search, structured metadata, and access-control filters; determine candidate depth; rank documents and passages; and select precise evidence for the model context.
Ranking and Relevance
Define first-stage and later-stage ranking strategies using lexical relevance, embedding similarity, freshness, behavioral signals, business rules, tensor features, and learned models. Evaluate candidate recall, ranking quality, latency, and inference cost across the query path.
Personalization and Recommendation
Model user, item, and contextual features; incorporate real-time behavioral signals; retrieve candidates; and rank recommendations, feeds, search results, and next-best actions at query time.
Production Serving
Design query and feed paths around corpus size, QPS, update rate, freshness, latency, availability, deployment, observability, and infrastructure cost. Evaluate when to use indexed search, streaming search, or a combination of both.
About Jenny
I’m Jenny Morris, a Senior Principal Solutions Architect at Vespa.ai. I work with engineering teams to design and evaluate production systems for AI search, RAG, ranking, recommendation, and personalization.
Before moving into solutions architecture, I designed and developed mission-critical production applications. My background spans computer science, chemical engineering, distributed systems, search, databases, and cloud-native infrastructure, with technical roles at Vespa.ai, Weaviate, Elastic, Pivotal, and Oracle.
I hold master’s degrees in Computer Science and Chemical Engineering from Washington University in St. Louis.
I also understand how complex infrastructure is evaluated and adopted inside large organizations. That perspective helps me design architectures that are not only technically sound, but also operable, governable, and supportable in production.
Case Studies
Perplexity uses Vespa.ai to power fast, accurate, and trusted answers for millions of users.
Yahoo has leveraged Vespa Cloud to build most of their highly scaled personalized interactive experiences.
Spotify relies on Vespa for a variety of use cases.