We're thrilled to announce our partnership with Albert School 👩🎓 🔗 Albert School is creating the next generation of digital transformation pioneers and leaders, through its unique approach to business and data education. Through this partnership, Linkup will provide Albert School students and researchers with custom access to its state-of-the-art search API for AIs, empowering them to build innovative AI agents serving real-world business use cases. Albert School and Linkup are together pushing the boundaries of knowledge and accelerating innovation in education. cc Grégoire Genest
Linkup
Technologie, information et Internet
Linkup is the web retriever that superpowers products & agents
À propos
At Linkup, our mission is to connect AIs to the internet. We build the infrastructure for AIs to access the world’s information efficiently, legally and at a fair price.
- Site web
-
http://www.linkup.so
Lien externe pour Linkup
- Secteur
- Technologie, information et Internet
- Taille de l’entreprise
- 11-50 employés
- Siège social
- Paris
- Type
- Société civile/Société commerciale/Autres types de sociétés
- Fondée en
- 2024
Lieux
-
Principal
Paris, FR
-
The Bay
San Francisco, California, US
Employés chez Linkup
-
Laurent Sarfati
Sales Coach | Ex Head of Sales @Qonto 🦄 & @Doctolib 🦄 #Next40
-
Guillaume Larcher
Freelance conception logicielle et IA 🤖 Renfort ponctuel | Accompagnement sur le long terme
-
Ayoub Assabban
Proudly taking over Assabban SA & Early Stage Investor l Advisor (cal.com/assabban)
-
Raphaël Escrig
Co-fondateur & CTO Newgo - Ingénieur Full Stack chez Linkup
Nouvelles
-
🇧🇷
Dia sim, dia também, enfim um estruturado Corrective RAG construído! Mais uma vez, mais uma vez inspirado diariamente em Ashpreet B., Martin Khristi, Lester Oliver, Eric Ciarla, Avi Chawla, Adil Aziz, Akshay Pachaar e Shubham Saboo. A interface apresenta claramente quatro indicadores de status fundamentais: Google API, Firecrawl, PyMuPDF e PDFNumber, todos devidamente marcados como funcionais, estabelecendo assim a base tecnológica necessária para as operações de recuperação e geração aumentada. O sistema é estruturado com Docling para parsing de documentos complexos, proporcionando capacidade superior de interpretação estrutural mesmo em textos com formatação intrincada, enquanto a integração permite a ingestão eficiente de múltiplos formatos de dados tais quais JSON, CSV, PDF, conjuntos de dados Kaggle e conteúdos via URL. O painel esquerdo expõe as configurações críticas do sistema, começando com o método de orquestração RAG que demonstra três opções distintas: None (atualmente selecionado), Gemini Direct e Agno with Gemini. A flexibilidade de orquestração permite ajustar o comportamento do sistema conforme as necessidades específicas de cada caso de uso também expondo a seção de RAG Mode, onde três abordagens metodológicas estão disponíveis: RAG Debug (ativamente selecionado), Corrective RAG e Combined Mode, oferecendo diferentes estratégias para o refinamento de respostas baseadas em documentos! Ademais, há uma seção dedicada à configuração da Linkup API, explicitamente marcada como requisito para o CRAG, com campo para inserção da chave API e link para obtenção da mesma. O sistema é alimentado pelo framework LlamaIndex, conforme indicado, estabelecendo assim uma base sólida para operações de indexação e recuperação semântica, com a arquitetura permitindo definir limites de similaridade personalizados (tipicamente 0.65) e utiliza conhecimento geral quando informações específicas não são encontradas nos documentos processados. O processamento de documentos atinge capacidades impressionantes, com suporte nativo a arquivos em até 200MB e segmentando/indexação avançadas mesmo em arquivos complexos com múltiplas camadas. O pipeline de refinamento de dados incorpora Gitingest para versionamento, Abacus.AI para normalização estrutural e integrações Qodo (formerly Codium)/Cline/Roo Code com embeddings Gemini 03-07, incluindo personalização de idioma para português ou qualquer outra língua necessária. A refinar sempre e sempre!
-
-
Linkup a republié ceci
At SOTA, we're launching a series of deep dives into AI topics, designed for engineers, tech founders, and AI researchers who want to go beyond the surface. For our very first edition, we’re tackling a hot topic, the Model Context Protocol by Anthropic, especially since its adoption by Cursor. To break it down and explore its implications, we’ve assembled 4 top-tier experts: 🔥 Louis Abraham from H Company, 🔥 Philippe Mizrahi from Linkup, 🔥 Thomas Payet from Meilisearch, 🔥 Raoul Ritter. They’ll dig into what MCP changes, debate its strengths and limitations, and share their best predictions for its future. Expect a sharp, insightful discussion, followed by an interactive Q&A. And that’s just the beginning! Soon, we’ll also be hosting Laurent Mazare from Kyutai, one of the godfathers of our SOTA community. 📅 The Live Webinar will happen on Contrast 🕔 17h00 Paris — 3 PM London — 11 AM SF — 8 AM NYC 👉👉👉 Register here: https://lnkd.in/erTBZ9i7 👈👈👈 Join us and be part of the conversation! It’s free, open to all, and built on collaboration. (Spice level: 🌶️ - builders’ empowerment, zero fluff, easy to follow if your a tech)
-
-
Linkup a republié ceci
I've built a "Peer Finder" using Linkup API. Context: you have 3-5 companies in mind and you want to find peers (competitors or other firms in the ecosystem). How does it work? ▶️ Ask Linkup to define the common characteristics of these companies. ▶️ Linkup proposes the common characteristics. ▶️ Choose the ones you want to retain (or you can even adjust them, e.g., adding criteria such as "VC-backed"). ▶️ Ask Linkup to search for companies that fit these criteria. Guess what? It works with people too! (marketers out there: it defines your persona and finds 50, 100, 200... other profiles 😉 ) Tell me if you're interested in the script (please specify whether it's people or companies you're looking for) before we release a complete tutorial.
-
Linkup a republié ceci
I just build a corrective RAG (CRAG) agentic workflow! It's powered by a locally running DeepSeek-R1 and has ability to search through your docs, evaluate the context quality and fallback to web search if it needs more info. Tech stack: - LlamaIndex workflows for orchestration - Linkup for deep web search - Comet's Opik to trace and monitor - Qdrant to self-host vectorDB The video below shows a quick demo of how the app works! Here's an overview of what the app does: — First search the docs with user query — Evaluate if the retrieved context is relevant using LLM — Only keep the relevant context — Do web search if needed — Aggregate the context & generate response I have published this work as a Lightning AI⚡️ Studio. It's reads like a blog encapsulating all my code & environment! Link in the comments! _____ Find me → Akshay Pachaar ✔️ For daily insights and tutorials on LLMs, RAG and AI Agents.
-
What does a debate between Plato and Aristotle have to do with what we do at Linkup? More than you’d think... Read more about how ancient thoughts influence the debate between Symbolic Reasoning and Machine Learning AI in our latest blog post.
What does a debate between Plato and Aristotle have to do with Linkup? Quite a lot, actually. Plato saw knowledge as the discovery of eternal truths. Aristotle believed it emerged from experience. AI development is echoing this same debate: should intelligence be built from structured logic or uncovered from data? We believe the answer lies in both. By combining symbolic systems (Platonic knowledge) with machine learning (Aristotelian experience), we’re building AI that retrieves, reasons, and refines - delivering intelligence with the precision of an expert. The road to truly intelligent AI runs through Athens. We’re walking the way. Read more about this fascinating debate in our lastest blogpost. Link in the first comment.
-
🚀 Powering Fleet's AI Copilot with live web search We're thrilled that Linkup's search API is part of Fleet's AI Copilot. While most AI assistants are limited to their knowledge base, Fleet's Copilot can access real-time information across the web thanks to our integration. IT teams can get precise answers to technical questions from sources they can trust. What makes this integration special? Ultra-low latency, intelligent source filtering, and response formats specifically designed for AI agents - all working together to enhance Fleet's users support experience. We're pleased to see Fleet's success, recently featured as Product of the Day on Product Hunt. Interested in the full story? Read our blog post here: https://lnkd.in/eyhVthRq If you’re building an AI copilot (or any agent) and need trusted, structured, and highly relevant access to web data, we’d love to talk!
-
-
Introducing our new Pay-As-You-Go pricing: Simplicity meets transparency We've heard our customers' feedback: pricing should be simple and transparent. That's why we've removed subscriptions entirely in favor of a pay-as-you-go model. What it means for you: ▶️ No more credits concept ▶️ No monthly engagement ▶️ No token-based pricing, for straightforward cost predictions How does it work? ✅ Charge your wallet and pay only for what you use ✅ Opt-in to auto-recharge for uninterrupted service ✅ For enterprise customers: contact us for tailored plans that fit your specific needs Sign-up here: app.linkup.so
-
Linkup a republié ceci
Linkup just set a new state-of-the-art performance on SimpleQA (OpenAI's factuality benchmark for AIs) at 90.10%, beating Perplexity and the rest. Linkup grounds AIs on high quality data to make them more accurate. With Linkup, GPT-4o mini's accuracy jumps 10x (!!) on SimpleQA. Read about it in our blog (link in the first comment). cc Boris Toledano Denis Charrier Guillaume Larcher Clément Jumel Raphaël Escrig Sacha Uzan Juliette Sivan
-
Linkup a republié ceci
A friend working in the luxury shoe industry asked for my help to urgently gather information about her competitors. She provided me with an Excel spreadsheet containing a list of companies and the data fields she needed. We used the Linkup API directly in Google sheets and created the *Linkup_Search* function that takes the row and column titles as input. 200 fields filled in seconds. The *Linkup_Search* function is available in our documentation (link in comments). Otherwise, just send me your Excel file and I'll fill it for you 😌. cc Philippe Denis
Pages similaires
Parcourir les offres d’emploi
Financement
Dernier round
Amorce de mise de fonds initiale3 178 363,00 $US