enki GmbH

Wird 2025 das Jahr der Agents? 🤖 Die Nützlichkeit von Large Language Models (LLMs) hängt stark von ihren Umsystemen ab. Ein Trend, der immer mehr Aufmerksamkeit erhält: Agents. Die Kernidee ist, dass mehrere spezialisierte LLMs zusammenarbeiten – z. B. um Fragen zu analysieren, relevante Antworten zu sortieren und dann die beste Antwort zu formulieren. Oft ist das Ergebnis deutlich besser als mit bisherigen Ansätzen. Auch wir bei enki haben uns mit Agents beschäftigt. Die Vorteile sind spannend, aber sie kommen mit einer erhöhten Komplexität. Nützlichkeit hat eben ihren Preis. Egal, ob Agents 2025 durchstarten oder nicht – wir werden auch im neuen Jahr unseren Kunden helfen, diese Komplexität zu zähmen und mit ihnen zusammen nützliche Systeme bauen. Einen grossartigen Start ins neue Jahr 🚀

enki from now on in German! Vor einiger Zeit wurden wir gefragt, warum unsere Kommunikation auf Englisch stattfindet, obwohl unsere gesamte Zielgruppe in der Deutschschweiz ansässig ist. Unsere erste Antwort war: Nun ja, ähhh… weil… Computer und so. Das war natürlich nicht optimal. 😉 Daher haben wir unsere Website nicht nur ins Deutsche übersetzt, sondern auch grundlegend überarbeitet. 👉 Das Ergebnis seht ihr hier: https://www.enki.swiss Wir sind stolz darauf, nun Teil der Swiss Made Software und des Swiss Digital Services Labels zu sein. Damit bekennen wir uns zu Schweizer Produkten und Dienstleistungen – und zu Werten wie Zuverlässigkeit, Integrität, Innovation und Qualität. Es ist uns ein Ansporn diesen Werten gerecht zu werden.

🚀 Why We Focus on Custom ML Systems Instead of Serving LLMs as a Service We often get asked why we don’t offer LLMs as a service. The answer is simple: economics. To match the level of a model like GPT-4o-mini, we would need something like LLaMA 3.1 70B. On a single Nvidia H100 machine, we could serve around 20 tokens per second. In Switzerland, 1 million tokens would cost us about $137. Even with optimized usage and serving multiple users, the cost would still be around $6.8 per 1M tokens. Compare this to OpenAI's API, where 1 million tokens cost just $0.26—a staggering 26x cheaper. Competing on price for the same product is nearly impossible, given the scale and efficiency of providers like OpenAI. That's why we specialize in custom and smaller LLMs. If you have a specific use case or prefer not to send your data to OpenAI, we’re here to offer tailored LLM solutions in Switzerland—just not at rock-bottom prices. 😄 #enki #LLM

🚀 Exciting news from Apple this week as they unveiled their AI strategy Apple's strategy involves using a compact base model, intricately fine-tuned through numerous LoRA adapters, each tailored for specialized tasks. This approach not only enhances efficiency but also sharpens the effectiveness of AI applications. 🔍 This strategy is a clear indicator that even industry giants like Apple are steering away from the rush towards AGI (Artificial General Intelligence). Instead, they're capitalizing on specialized models that excel in specific domains. At enki, we resonate with this philosophy! We specialize in crafting customized AI solutions tailored to your unique business needs. Our models are designed to integrate seamlessly into your operations, driving innovation and creating business value. 🌟 Let's innovate together and transform your business with AI solutions that are as unique as your challenges! Source: https://lnkd.in/gJgjCzzG #ML #enki #AppleAI

🚀 Yes we RAG! We’ve built a powerful Chatbot that knows everything about The Simpsons with 16,000 documents feeding its knowledge base! And the best part? It’s all running on our infrastructure – no external services like OpenAI involved! 🤖🔍 Imagine, instead of chatting with a bot that's an expert on The Simpsons, engaging in seamless conversations with a Chatbot that holds the collective knowledge of your company. This tool can help you develop new ideas, effortlessly search for documents, and enhance overall productivity. Curious to have your own specialized Chatbot? Reach out to us! 💬 #enki #RAG

🚀 Small ML models are incredibly exciting! While Large Language Models (LLMs) have their place, sometimes smaller, purpose-built models reign supreme. Take a peek into one of our dataset creation pipelines: each purple box represents a unique small ML model designed for a specific task. These models are faster, more cost-effective, and crucially, more precise than any LLM to date (at least for now 😉). #enki #ML #SmallModels

Introducing Sketch2Img - Our Newest Model on the Farm! 🚀🎨 Are you constantly sketching to showcase ideas or impress clients? Say hello to Sketch2Img, your ultimate tool for lightning-fast iteration and visualization! With Sketch2Img, effortlessly create diverse variants of your sketches, each with unique styles and surroundings. Whether you're designing buildings, interiors, or products, this innovative model accelerates your creative process. Curious for a demo or have questions about its capabilities? Reach out to us today! #Sketch2Img #enki