MLOps: Deploying LSTM Model for Career Prediction

I designed and implemented an MLOps process for a data science team deploying their LSTM-based career prediction model. While the team focused on model development, I created an efficient and scalable deployment solution tailored to their needs. Initially, I evaluated Google Vertex AI, but its costs proved prohibitive for this MVP deployment.

Instead, I implemented a cost-effective alternative using a Google Compute Engine VM running a TensorFlow Serving container. Predictions were served via a REST API through a Google Cloud Function, offering seamless integration. To further enhance the workflow, I automated the deployment process, allowing model updates to be tested in minutes after being pushed. This solution minimized costs while maximizing flexibility, enabling rapid iteration and reliable deployment without compromising on quality or scalability.