Mohamed A.

Design and implementation of a scalable architecture for a client project on Upwork, using AWS Lambda, SageMaker, and FastAPI. Utilization of an Application Load Balancer to efficiently manage traffic and ensure no request loss. Integration of Google OAuth for secure user authentication. Achieved dynamic scaling of Lambda functions based on request volume. Ensured seamless connectivity between Lambda functions and the SageMaker endpoint. Resulting in efficient traffic management, dynamic scalability, and secure user authentication.

Complete design and development of an innovative recommendation application facilitating collaboration between startups and investors.

​

● Utilization of Amazon EMR on AWS with PySpark for data processing.

● Storage of processed data in Amazon S3.

● Development of models on data samples.

● Creation of jobs on Amazon SageMaker for model training.

● Creation of pipelines to orchestrate the model development process.

● Use of pipelines and unit tests to ensure code quality.

● Commit of the project to AWS CodeCommit.

● Deployment of the saved model on Amazon SageMaker using auto-scaling capabilities and adding an autoscale endpoint, a Lambda function, and an API Gateway for API development.

● Execution of thorough tests in the Test environment to validate the application's proper functioning.

● Approval of deployment after test validation.

● Deployment in the production environment using best practices for security and scalability.

● Utilization of Amazon CloudWatch to monitor model performance and detect changes in data distribution.

● Capture of inference data to track the evolution of results.

Audio Data Collection and Labeling:Build a database of audio recordings with accurate emotional annotations to train the emotion recognition model.

●Audio Signal Transformation:Apply transformations to audio signals, such as calculating fundamental frequency, Mel-frequency cepstral coefficients (MFCCs), etc., to extract relevant features for emotion recognition.

●Deep Learning Model Training:Use deep neural networks to learn to classify emotions from features extracted from audio signals.