Posts

Our team was selected to participate in the “Solve for Healthcare & Life Sciences with Gemma” hackathon, hosted by Google Paris! I had the pleasure of teaming up with my former ETH Zürich lab-mate Patrick Langer, along with Nour Ben Rejeb and Nathan Denier. We were honored to receive the Impact Award (2nd place) among 126 participants across 26 teams, a great recognition of how our research at ETH Zürich and Stanford University translates into real-world impact. ...

Generative AI is revolutionizing the way we work, enabling new levels of efficiency and creativity. At AWS, we understand that many organizations rely on Microsoft Office applications like Word, Excel, and Outlook. That’s why we’ve developed a solution that seamlessly integrates AWS generative AI capabilities, powered by Amazon Bedrock, into these familiar tools. This integration allows employees to leverage large language models (LLMs) directly within their daily workflows, enhancing productivity and insights. ...

In the rapidly evolving landscape of data analytics, leveraging generative AI can significantly enhance efficiency and insights derived from data. The No-Code Generative AI Data Analytics Solution on AWS 🧠📊 offers a compelling approach to integrating AI-driven analytics without requiring extensive coding knowledge. This solution empowers users to transform their data analysis workflows by enabling natural language prompts to generate and execute code automatically. Your browser does not support the video tag. At the core of this solution is the intelligent code interpreter 🤖, which utilizes Amazon Nova foundation models on Amazon Bedrock for sophisticated reasoning capabilities. The entire setup is securely hosted within an Amazon Virtual Private Cloud (VPC) 🔒, ensuring enterprise-grade security. The architecture is built using AWS serverless components like Lambda ⚡, API Gateway 🌐, and DynamoDB 🗄️, allowing for scalable and efficient handling of requests. The user interface, based on Streamlit, is deployed on Amazon ECS Fargate 🚢, providing an intuitive way for users to interact with the solution. ...

To help fellow Generative AI enthusiasts quickly leverage cutting-edge open-source models, I’ve created a simple CLI tool that bridges Hugging Face and Amazon Bedrock. The Hugging Face Bedrock Importer is a command-line tool that simplifies the process of downloading Hugging Face models and deploying them to Amazon Bedrock. This tool automates the workflow of model download, Amazon S3 upload, and Bedrock Custom Model Import, making it easier for developers to leverage Hugging Face language models in their AWS environment. ...

Discover how BMW Group, in collaboration with Boston Consulting Group (BCG) and Amazon Web Services (AWS), is transforming procurement with advanced generative AI. The Offer Analyst tool streamlines offer evaluations, improving efficiency and accuracy. By automating tedious tasks and reducing manual effort, this innovative solution addresses key pain points in traditional procurement workflows. Learn how Offer Analyst is setting new standards in procurement excellence and delivering significant business impact. ➡ Read more about BMW Group’s Offer Analyst in our AWS blog post. ...

Embarking on the journey to obtain all AWS certifications has been both challenging and rewarding. Here’s a breakdown of my certification path: Foundational and Associate Certifications With some general AWS experience under my belt, I began with the foundational and associate-level certifications: AWS Certified Cloud Practitioner AWS Certified Solutions Architect – Associate AWS Certified Developer – Associate Specializing in Machine Learning and Data Analytics After working for several months with AWS’s machine learning services, I pursued certifications aligned with my daily work: ...

Discover how to leverage AWS services to create a Retrieval-Augmented Generation (RAG)-enabled generative AI application. This solution integrates LangChain and Amazon Aurora PostgreSQL-Compatible as a vector store, enabling near real-time document ingestion. Architecture Overview This solution consists of several key components. Amazon S3 is used for file uploads, while AWS Lambda handles data processing. For model integration, the architecture leverages Amazon Bedrock. Workflow and architecture components ...

BMW Group has introduced the Data Interpreter, a groundbreaking solution that leverages generative AI to democratize data analytics. This no-code application empowers employees across various departments to perform complex data analyses using natural language instructions—without writing a single line of code. The solution utilizes Amazon Bedrock’s large language models and the ReAct framework for iterative code generation, ensuring accurate and efficient data interpretation. ➡ Learn more about our solution in this AWS blog post. ...

We recently published a paper at the 2023 CHI Conference on Human Factors in Computing Systems, presenting an in-vehicle machine learning system designed to predict critical blood alcohol concentration (BAC) levels in real-time. This system leverages driver monitoring cameras—now mandated in many countries—to assess a driver’s BAC. Our evaluation involved a simulator study with 30 participants, demonstrating that the system reliably detected any alcohol influence while driving with an area under the ROC-curve (AUROC) of 0.88. It also identified drivers exceeding the WHO-recommended limit of 0.05 g/dL BAC with an AUROC of 0.79. Model inspection revealed that the system relies on pathophysiological effects associated with alcohol consumption. ...

Exciting advancements are underway in the field of diabetes management with the introduction of HEADWIND, an advanced in-vehicle hypoglycemia warning system. This cutting-edge technology, funded by the Swiss National Science Foundation (SNF), aims to detect hypoglycemia in drivers with diabetes early and accurately by utilizing existing automotive sensors and data. The system integrates high-resolution, real-time driving data, physiological sensors, and driver camera video streams for eye and head pose tracking to identify hypoglycemic episodes. ...