Skip to main content

Atita Arora

she/her

Building search, RAG & agentic systems for 18+ years | Apache OpenNLP PMC Member | Co-lead, Women of Search

📍 Germany 💼 Bistro Innovation Labs GmbH

About Atita

Atita Arora is a hands-on search and AI engineer whose work runs from enterprise search and NLP to vector search, RAG, and agentic systems. She has built and scaled search at Idealo and MYTOYS Group, and worked as a Solutions Architect at Qdrant.

In open source, Atita is a PMC member of Apache OpenNLP and serves on the program committee for Berlin Buzzwords. She co-leads Women of Search, a community for women working in AI, search, and agents, where she runs events and helps new speakers get their first stage.

She is a regular speaker at Haystack and Berlin Buzzwords, and writes on VectorHub, including a three-part series on RAG evaluation. Her current focus: why the search community’s decades of relevance expertise are exactly what today’s AI agents and their memory systems are missing.

You can also hear her on the Vector Podcast talking vector search for e-commerce. More slides are on Speaker Deck.

Atita is based in Berlin.

Speaking Topics

Your Agents Are Only As Good As Your Retrieval

When agents can issue refunds, cancel accounts, or delete data, a retrieval miss stops being a bad answer and becomes a wrong action. With a live notebook demo, this talk shows how an agent can follow its instructions perfectly and still act on the wrong document, why context engineering is relevance engineering for a new user, and why you should evaluate retrieval before you orchestrate. Watch the talk · Slides · Demo notebook

Previously delivered at:
  • Context Camp by Vanishing Gradients, 2026

You're Solving the Wrong Problem with AI Search

Teams rebuild relevance when the real issue is the catalog, add vector search when it's the schema, and tune ranking with no signal to tune on. This session lays out a diagnostic framework of about ten recurring failure patterns across e-commerce, media, legal, finance, and geospatial search, and shows why agentic retrieval inherits every one of them. Watch the session · Slides

Previously delivered at:
  • The Frontier of AI Search, Maven lightning lesson series hosted by Trey Grainger and Doug Turnbull, 2026

AI Agents: From Hype to Reality

A Women of Search session on what agentic AI can and can't do in production today, and how to separate real capability from demo magic. Co-presented with Angeley Mullins and Olena Gorbatiuk. Watch the talk · Slides

Previously delivered at:
  • Haystack EU 2025

From Reactive to Proactive: Is Agentic AI the Future of Decision-Making?

How agentic systems shift AI from answering questions to taking initiative, and what that means for the products and decisions built on top of them. Watch the talk · Slides

Previously delivered at:
  • Product-Led Hub 2025

Navigating RAG Optimization with an Evaluation-Driven Compass

Ten experiments on a documentation Q&A system show how to tell missing evidence apart from noisy context, then fix each failure with the right tool, from reranking to hybrid search. Co-presented with Deanna Emery. Watch the talk · Slides

Previously delivered at:
  • AI Engineer World's Fair 2024

Cracking the Code: Deciphering Evaluation Essentials for RAG

The core components of a RAG evaluation setup, how existing tools and frameworks compare, and how to stay in control while you improve your system. Watch the talk · Slides

Previously delivered at:
  • Berlin Buzzwords 2024

Vectorize Your Open Source Search Engine

A practical starting point for adding vector search to the open source engine you already run. Watch at Berlin Buzzwords · Watch the e-commerce edition in Kraków · Slides

Previously delivered at:
  • Berlin Buzzwords 2023
  • Haystack on Tour, Kraków 2023 (as "Vectorize your E-commerce Search")

Women of Search

Sessions from the Women of Search community at Haystack, putting women's voices and work in search on the main stage. Watch 2023 · Watch 2022

Previously delivered at:
  • Haystack EU 2023 (with Paige Tyrrell and Istvan Simon)
  • Haystack EU 2022

Understanding Vespa with a Lucene Mindset

An introduction to Vespa's architecture and advanced features for engineers who already know Lucene-based engines like Solr and Elasticsearch. Watch the talk

Previously delivered at:
  • Berlin Buzzwords 2022

Exploring the Alchemy of Streaming and SolrCloud

How an e-commerce platform moved from four-hour batch reindexing to near-real-time streaming with Kafka and Redis, and migrated from master-slave Solr to SolrCloud along the way. Watch the talk

Previously delivered at:
  • Berlin Buzzwords 2021