Build and Deploy Retrieval Augmented Generative Pipelines with Haystack

Retrieval Augmented Generative pipelines (RAG) are one of the most effective techniques to leveraging LLMs for production.

In this talk, Tuana Celik will walk through how you can build and customize your own RAG pipelines with Haystack, the open-source LLM framework by deepset. She will demonstrate the Haystack pipeline structure, which is built on the principle of composability, and how you may shape custom pipelines to:

  • Make use of the vector databases of your choice
  • Shape out how your data will interact with LLMs
  • Use the latest retrieval techniques to provide LLMs with the relevant context  and more.

Following that, Thomas will discuss the challenges of deploying such pipelines to production.

  • You will learn how we deploy Haystack pipelines on deepset Cloud. In particular we’ll address:
  • Hosting a pipeline
  • Connecting your pipeline to LLMs hosted on AWS SageMaker
  • What it actually means to run a RAG pipeline end-to-end covering prototyping, evaluation, inference, prompt engineering, observability and more


Thomas Stadelmann

Senior Software Engineer at Deepset

Thomas Stadelmann is Senior Software Engineer at Deepset, where he works on Haystack, Deepset's Open Source Framework for Retrieval Augmented Generation, and on deepset Cloud, their complementary commercial product. Thomas joined Deepset in 2021. Coming from the field of Information Retrieval, he has been working on real-world search applications for over 12 years.

Read more


Tuana Celik

Developer Advocate @Deepset

Tuana works as a Developer Advocate for Haystack, the open source LLM framework by Deepset. She helps other engineers and the NLP community understand the latest approaches to build LLM applications, and in particular how they can do so with Haystack.

Read more

Session Sponsored By

deepset offers enterprise developer tools to build NLP-driven applications using LLMs.


Monday Oct 2 / 01:35PM PDT ( 50 minutes )


Pacific LM


From the same track


The 10 REST Commandments

Monday Oct 2 / 11:45AM PDT

Learn how to build amazing REST-based APIs that your customers will love with ten universally applicable and easy-to-follow rules. The rules are: 

Speaker image - Vedran  Cindrić

Vedran Cindrić

Founder and CEO @ Treblle


Delegate Engineering Toil to the Robots

Monday Oct 2 / 03:55PM PDT

Developer Experience is Real. And it’s not defined by metrics. It’s defined by the ability to get rid of all the parts of the job that suck, and empowering your developers to focus on their core job. DevEx is about removing the suck.

Speaker image - Dylan  Etkin

Dylan Etkin

CEO & Co-Founder @Sleuth


Rethinking SaaS Entitlement Management with Feature Flags

Monday Oct 2 / 10:35AM PDT

Feature flags are typically associated with continuous delivery feature access management. They allow you to temporarily limit production access to a new feature while it's being developed and tested, then open access for your users once it's ready.

Speaker image - Brian Rinaldi

Brian Rinaldi

Developer Experience Engineer @LaunchDarkly

Speaker image - Bill Tarr

Bill Tarr

SaaS Evangelist, AWS SaaS Factory @AWS


Beyond the Hype: The Impact of GenAI on Software Engineering Teams Productivity

Monday Oct 2 / 02:45PM PDT

In a realm often saturated with predictable discussions about artificial intelligence, StackSpot is pioneering a fresh approach to Gen AI.

Speaker image - João Batista Neto

João Batista Neto

Product Manager @StackSpot


Building Organizational Resilience by Creating Generative Organizations

Monday Oct 2 / 05:05PM PDT

2004, Professor Ron Westrum published his influential paper on how organizations process information. In that paper, Prof. Westrum described three archetypes of organizations, along with characteristics of how they process information. They are pathological, bureaucratic, and generative.

Speaker image - Rishi Markenday

Rishi Markenday

Engineering leader @McKinsey & Company