deepset, makers of Haystack | LinkedIn (original) (raw)

About us

deepset is the creator of the open-source Haystack AI orchestration framework and the Haystack Enterprise Platform. We help organizations build, run, and govern custom AI agents and retrieval-augmented generation (RAG) applications across cloud and on-premise environments. Teams start with Haystack to build immediately, add optional enterprise support through Haystack Enterprise Starter, and scale with the Haystack Enterprise Platform, which provides tooling for advanced indexing and retrieval, testing and evaluation, and governance for production AI. Haystack is trusted in production by enterprise teams and organizations including the European Commission, The Economist, Airbus, and the German Armed Forces.

Industry

Software Development

Company size

51-200 employees

Headquarters

Berlin, BE

Type

Privately Held

Specialties

Artificial Intelligence, Natural Language Processing, Question Answering, Large Language Models, AI, LLMs, Machine Learning, Semantic Search, Conversational AI, Question Answering, Information Extraction, vetor-based search, and Document Similarity

Locations

Employees at deepset, makers of Haystack

Updates

deepset, makers of Haystack reposted this
🇩🇪 Great news out of the European tech ecosystem! As announced at the #ProcessIntelligenceDay in Berlin today Celonis and deepset, makers of Haystack are joining forces to build a sovereign AI platform designed for mission-critical operations. In highly sensitive sectors like defense, cybersecurity, and the public sector, trust and control are non-negotiable. That is exactly what this partnership delivers. By combining Celonis' powerhouse Process Intelligence with deepset's governed AI agent framework (Haystack), organizations no longer have to choose between cutting-edge AI and strict data sovereignty. The result? Grounded, highly auditable AI systems that understand real-world operational context—giving teams full control over their models, infrastructure, and data without vendor lock-in. One of the many right steps for European digital sovereignty! More info in the comment section 👇
deepset, makers of Haystack reposted this
AI for the public sector and critical infrastructures has special requirements. It works best with domain-specific (often sensitive) context and clear guardrails. When it comes to key components, like language models, and compute, sovereignty is not optional anymore. We need technology-agnostic frameworks, such as Haystack AI Framework. This became undeniably clear over the weekend: The US government banned Anthropic AI-Models due to "national security" concerns. Export controls now apply to models, not just chips. That's why European partnership matters more than ever. This is exactly, why I am excited to announce our partnership with Celonis that we officially launched yesterday. We've been working closely with Julius Ibel, Oliver Zeller and the team in the past months to build a reliable solution "made in Europe" by combining our strenghts. Looking forward to whats ahead. Link to full statement in the comments. #sovereignai #publicsector #aigovernance #madeineurope #deepset

``
23,950 followers
22h Edited
Sicherheit und Souveränität entscheiden darüber, ob agentische KI in sensiblen Umgebungen überhaupt in Produktion geht. Am Mittwoch zeigen wir, wie das gelingt. Gemeinsam mit Hewlett Packard Enterprise, NVIDIA und secunet Security Networks AG stellen wir eine Referenzarchitektur vor. Sie läuft in geschlossenen Netzen, ist VSA-konform und lässt sich in Wochen umsetzen. Worum es geht: • Wie die vier Layer ineinandergreifen: secunet SINA Cloud, HPE Private Cloud AI, NVIDIA NIM und deepset Haystack • Warum reines Abschotten der Netze nicht ausreicht AIDA als konkreter Use-Case: ein Multi-Agenten-System zur Entscheidungsunterstützung • Wie Sie die Architektur auf Ihre eigene Umgebung übertragen Mit dabei: Hartmut Schultze (Hewlett Packard Enterprise), Dr. Robert Friedrich (NVIDIA), Jens Plogsties (SysEleven GmbH) und unsere Dr. Anna Gründler (deepset, makers of Haystack). 📅 Mittwoch, 17. Juni, 14:00 bis 15:00 CEST, virtuell Für IT-Architektinnen und -Architekten, Plattform- und Infrastrukturverantwortliche und KI-Entscheider in regulierten Umgebungen. Anmeldung: https://lnkd.in/eKntkVDqSecurity and sovereignty are key factors in determining whether agentic AI will ever go into production in sensitive environments. On Wednesday, we will demonstrate how this can be achieved. Together with Hewlett Packard Enterprise, NVIDIA and secunet Security Networks AG, we will present a reference architecture. It runs on closed networks, is VSA-compliant and can be implemented within weeks. What it’s all about: • How the four layers interlock: secunet SINA Cloud, HPE Private Cloud AI, NVIDIA NIM and deepset Haystack • Why simply isolating networks is not enough • AIDA as a concrete use case: a multi-agent system for decision support • How to adapt the architecture to your own environment Speakers: Hartmut Schultze (Hewlett Packard Enterprise), Dr. Robert Friedrich (NVIDIA), Jens Plogsties (SysEleven GmbH) and our own Dr. Anna Gründler. 📅 Wednesday, 17 June, 14:00 to 15:00 CEST, virtual Register here: https://lnkd.in/eKntkVDq
Today we're announcing a partnership with Celonis to build an integrated sovereign AI platform for mission-critical operations. With this partnership, we aim to bring trustworthy, auditable, and sovereign AI to the organizations that need it most: public sector, defense, cybersecurity, policing, and critical infrastructure. This partnership combines: • Celonis' operational intelligence and the new Celonis Context Model (CCM), which gives AI systems a real-time understanding of how an organization actually operates • deepset's governed platform for AI agents, built on Haystack AI Framework, for sovereign infrastructures where transparency, control, and auditability are essential requirements The result is a joint platform that enables teams to unify fragmented data, detect operational risks earlier, and support coordinated decision-making. Teams keep full control: the platform runs on your preferred infrastructure, with no vendor lock-in. Together, we are creating a European alternative for operational AI that is transparent, auditable, and grounded in real-world operational context. Read the full announcement, linked in the comments.
From pilot to production, at scale, and on your own terms. At this year's Zukunftskongress in Berlin, our CEO Milos Rusic joined Timo Nink, Oliver Schmidt (Accenture) and Dr. Eva-Maria Hempe (NVIDIA) for a Best Practice dialogue on exactly that: how federal, state, and municipal organizations make the leap from isolated AI pilots to production-ready platforms. Three principles defined the conversation: • Sovereignty means tailored solutions, not all-or-nothing. Analyze your data, operational, and technology sovereignty requirements deliberately and design your architecture around what your organization actually needs. • Build for change, not from pilot to pilot. Your AI architecture has to survive the next technology leap. Interoperability across your application landscape isn't optional. It's what makes scale possible. • AI Orchestration is the core competency. Reusability is the condition for scale. Stop rebuilding from scratch. Wire together reusable AI components and microservices - built on an open core, designed for efficient operations in production. The panel's verdict: digital sovereignty is the baseline obligation for public sector organizations. Agentic AI and smart AI orchestration, that's where transformation actually happens. This is what Haystack AI Framework, deepset's open-source AI orchestration framework, is built for: giving teams the architectural foundation to move from proof-of-concept to real-world impact.#Zukunftskongress #AIOrchestration #SovereignAI #AgenticAI #Haystack #OpenSource #PublicSector #Accenture #NVIDIA
deepset, makers of Haystack reposted this
How do you build AI agents for classified domains? Next week, on June 17, 2026, 2–3 pm CET, I'll be presenting our answer together with Hartmut Schultze, Dr. Robert Friedrich, and Jens Plogsties: a Reference Architecture for agentic AI in classified environments. In the world of AI you can not win alone, you need to partner, build ecosystems so that everyone can bring their best to the table, in this case: deepset, makers of Haystack: Open-source AI orchestrationNVIDIA: Model inference, focused on open-source modelsHewlett Packard Enterprise: Private Cloud AI infrastructuresecunet Security Networks AG: BSI-certified security layer The result: a turnkey product to build ai agents and workflows in a way that meets the highest security standards. In the comments you find the registration Link for our joint Webinar. Heads up: The Webinar will be in German language - but the whitepaper is in English language, so please comment or reach out if you are interested to read it. Very much looking forward to see you at the Webinar. #agenticai #haystack #SovereignAI #SecureAI #AI #Cybersecurity #PublicSector #Haystack #NVIDIA #HPE #Secunet

``
deepset, makers of Haystack reposted this
Most AI frameworks are from outside Europe by default. Haystack isn't. Besides it's funky name it's built by deepset out of Berlin, it's an open-source Python framework for building production-grade AI agents and RAG pipelines. Modular components. Serializable pipelines. Full control over how context flows through your system. The customer list tells you it's serious: Airbus, The Economist, NVIDIA, the European Commission, the German Armed Forces, Siemens. What I like technically: you wrap your agent in a pipeline like any other component. That means your agent gets the same branching, looping, and conditional logic as your retrieval steps. No special-casing. No framework debt. For EU deployments it also matters that deepset ships a self-hosted enterprise option. Air-gapped if you need it. I've been evaluating it in products at Skarnlabs. So far it's the cleanest answer I've found to the "agent in a workflow" problem.#AI #OpenSource #AgentOps
AI governance should not slow innovation down. Done well, it gives teams the clarity and control they need to build AI systems that are trustworthy, scalable and ready for real-world deployment. Last Thursday, Luke Walker joined a panel discussion at IAPP AI Governance Global Europe on moving from theoretical compliance to operational AI governance. The session explored a tension many organizations are facing today: governance is often seen as a blocker, while technical teams are left to translate regulatory expectations into practical implementation. In reality, the biggest opportunity lies in better collaboration between legal, compliance and AI teams. A few key themes from the discussion: • The gap between policy and implementation, and how companies can bridge it • How structured governance creates clarity, trust and scalability • Why domain experts, not compliance teams or technology providers alone, should define what “good” AI output looks like • How transparency, testing and clear benchmarks help turn governance into an operational practice • Why AI compliance and AI sovereignty increasingly go hand in hand: both are ultimately about control over data, models, infrastructure and cost The EU AI Act arrives at a moment when enterprises are actively looking for guidance on how to adopt AI safely at scale. That makes this a critical opportunity to build governance approaches that enable innovation rather than constrain it. Thank you to Thomas-vini PIRES from EQS Group for hosting, and to Dr. Lasse Milinski from eagle lsp | Legal Tech Service Provider for the conversation. We were glad to contribute deepset’s perspective from the field and share what we are seeing as organizations move from AI experimentation to production.#AIGovernance #EUAIAct #EnterpriseAI #ResponsibleAI #AISovereignty #iappaigg26
deepset, makers of Haystack reposted this
This week I played Devil's Advocate on a panel at IAPP AI Governance Global Europe, the tech guy brought in to stir up the old Innovation vs. Regulation debate. Bit of a letdown, honestly. Turns out we mostly agree. 😅 Because IMO: • The EU AI Act is far better timed than GDPR, or cloud regulation before it. GDPR clamped down on business processes that had been running for years and cost economies billions. The EU AI Act shows up right when enterprises are craving guidance on how to do safe AI at scale. Totally different moment. • Compliance officers and tech providers actually agree on the thing that matters most: neither of us should decide what "good" output looks like. That's the domain experts' call. Our job on the tech side is to make AI systems transparent and easy to test. Compliance sets the benchmarks and makes sure they hold. • AI compliance and AI sovereignty? One hand washes the other. Both come down to control. Control over your data, your models, your infrastructure, your inventory. And that goes WAY beyond the public sector. AI gets more expensive by the day, and the same control that keeps you compliant is what keeps your bill predictable. Any CIO or CFO will tell you the same. Thank you to Thomas-vini PIRES from EQS Group for hosting, and to Dr. Lasse Milinski from eagle lsp | Legal Tech Service Provider for being a superb sparring partner on the legal side. Maybe next time we'll find more to disagree on. 😉 #iappaigg26 #EUAIAct #AIGovernance #SovereignAI

``

deepset, makers of Haystack | LinkedIn (original) (raw)

About us

Locations

Employees at deepset, makers of Haystack

Updates

Join now to see what you are missing

Affiliated pages

Similar pages

Browse jobs

Funding