Explore the cutting-edge capabilities of Vespa.ai, the open-source search engine pioneered by Yahoo, now a cornerstone for developers and data scientists seeking enhanced search and analytics functionalities. This session delves into the integration of machine learning to elevate search accuracy and user experience, achieving a fine balance between precision and recall. Participants will gain insight into Vespa.ai's evolution towards a cloud-based, analytics-focused platform, offering extensive configurability for Java aficionados and seamless incorporation of database-free text search capabilities.
The presentation further highlights the operational advantages and development efficiencies of Vespa.ai, including its compatibility with major big data technologies, automatic load distribution, and scalable deployment options. Attendees will leave with a comprehensive understanding of leveraging Vespa.ai for sophisticated, personalized search solutions within their own applications or platforms.
3. Search engine system, open sourced by
Yahoo in Sep 2017:
github.com/vespa-engine
Uses machine learning to provide much
higher quality search than alternatives.
Designed for low latency user interactions
and personalised results.
4. Provides search for Flickr, Yahoo news,
sports, finance and advertising.
(A 2010 commercial deal with Microsoft means
that Yahoo search uses Bing)
Derived from alltheweb.com codebase,
but refactored, modernised and
hardened. In production since 1999.
7. Focus has moved to
analytics
Cloud-based solution that
you pay for
8. Focus has moved to
analytics
Cloud-based solution that
you pay for
Highly configurable, if you
like Java
9. Focus has moved to
analytics
Cloud-based solution that
you pay for
Highly configurable, if you
like Java
Database free text search
functionality is "just there"
10. Focus has moved to
analytics
Cloud-based solution that
you pay for
Highly configurable, if you
like Java
Database free text search
functionality is "just there"
Easy to use machine
learning to improve results
15. How do they find
each other?
"Speakers for an internal seminar on
microservices in Birmingham"
"Conferences in Austin, Texas, about
Node.js, in early 2018"
"Audio/Visual service providers in
Brooklyn"
40. Most problems already solved -
batteries included philosophy
Development ❤
Well-defined extension mechanism
via Java OSGi plugins
Designed for integration with
Hadoop, Spark, Pig, etc.
41. All the fun of Java and XML
Development "
You really need to understand at
least the basics of machine
learning (tensors etc)
Really big system - significant
learning commitment
42. Automatically distributes load over
dynamically scaled hosts
Operations ❤
Official docker images make
deployment fast and simple
Full and deep documentation
43. No admin GUI - you make manual
requests against a REST API
Operations "
Resource hungry, especially
memory. You'll run multiple large
instances for production.
Very small community (as yet)