SING SYSTEMS
        Machine Intelligence – Cloud Engineering
       
      
        Services
        
          Dr. Tobias Sing is a visionary data/ML scientist with strong cloud
          engineering skills to turn ideas into solutions, with over 15 years of
          industry experience across various application domains. He is an AWS
          multi-certified expert in delivering secure, reliable, performant, and
          cost-optimized cloud systems, proficient in a wide variety of tools,
          libraries, and technologies. Dr. Sing is the author of popular
          open-source package ROCR for evaluating & testing ML models. His
          scientific excellence is demonstrated by
          over 30 peer-reviewed articles that have been cited in over 6,000+ publications.
        
        
          - 
            Robust Data Platforms & Pipelines. We architect and
            implement robust data platforms and efficient pipelines, ensuring
            seamless data ingestion, cleaning, and management to empower AI
            model development, enhance analytics, and optimize decision-making
            processes.
          
- 
            Next-Generation AI & ML Solutions. Harnessing the
            latest advancements in AI and machine learning, we develop
            customized models that drive innovation, enhance efficiency, and
            power transformative business solutions.
          
- 
            Cloud Deployment & Integration. From PoC/MVP to
            enterprise-grade: Transforming models and data pipelines into
            secure, reliable, high-performance, and cost-efficient applications.
          
          Please contact us if you would like to
          discuss a project.
        
      
      
        Interactive Engineering Showcases
        
          If you would like to explore our interactive showcases, please
          contact us. We will then provide you
          with the login credentials.
        
        
          - 
            Chat with your Documents: Retrieval-Augmented
                Generation.
            This demo showcases how Retrieval-Augmented Generation (RAG) can
            securely connect large language models (LLMs) with proprietary data.
            As an example, we created embeddings for the Bank for International
            Settlements (BIS) Annual Economic Reports (2018 onwards) using
            Amazon Bedrock’s Knowledge Bases and OpenSearch as a vector
            database, integrated with a Claude 3.5 Sonnet model for enhanced
            responses. The back-end is built with Python (deployed via AWS
            Lambda), API Gateway, and Bedrock, while the front-end consists of
            plain HTML/CSS/JavaScript, served efficiently and securely using
            Amazon CloudFront. Here are some sample queries, but feel free to
            ask your own questions:
            
              - 
                Are there recurring themes that consistently emerge in the BIS
                Annual Reports over the years?
              
- 
                In which report did BIS discuss rising oil prices in relation to
                the 1970s oil shocks, and what was the main takeaway?
              
- 
                How does BIS assess the economic impact of the global pandemic
                in its annual reports? Please summarize by year.
              
 
- 
            Economic Data Dashboard with Python and Plotly Dash.
            This dashboard is built with Python and Plotly Dash, containerized
            using Docker and deployed via AWS Fargate. In the back-end, a
            daily-triggered AWS Lambda function fetches new data from FRED and
            stores it in DynamoDB. The front-end allows users to interactively
            visualize economic indicators. Authentication is managed through AWS
            Cognito, and Elastic Load Balancing ensures seamless scalability.
          
- 
            Cost-Optimized Deployment of ML Models.
            AWS SageMaker provides a powerful platform for creating, training,
            and deploying machine learning models with seamless integration into
            the AWS ecosystem. This integration ensures high performance,
            availability, security, and cost-efficiency in production
            environments. Given the potential costs of model endpoints, this
            showcase highlights two deployment strategies—serverless and
            asynchronous—that minimize expenses by scaling down to zero during
            inactivity and scaling up on demand. Please allow a brief “warm up”
            time when accessing the endpoints after periods of inactivity.
          
- 
            LLM-based text summarization.
            A front-end developed with plain HTML/CSS/Javascript allows the user
            to submit a text, which is then summarized. The back-end consists of
            an API method deployed on Amazon API Gateway. The API Method calls a
            Lambda function which in turn invokes an LLM on Amazon Bedrock
            (Amazon Titan Text Express) to summarize the user-provided text.
          
Publications
        
          Dr. Sing's list of peer-reviewed scientific publications can be found
          on
          Google Scholar.
        
      
      
        Professional Certifications