Addressing Gaps and Weaknesses in Background Questions

Situation: After my last full-time role I had a planned one-year break to manage a family relocation and to focus on deliberate skill-upgrading in machine learning before re-entering production work.

Task: My goal was to stay current with SOTA ML practices and build production-ready artifacts that demonstrate I can design, train, and deploy models at scale.

Action:

• I created a self-directed curriculum: read top conferences (NeurIPS, ICML), followed transformer and MLOps developments, and completed targeted courses (DeepLearning.AI’s TensorFlow Dev, and an MLOps specialization).
• Hands-on practice: I built two concrete projects and contributed to open source to keep engineering rigor.

Result / Artifacts:

1. Real-time Anomaly Detection Microservice (GitHub repo): End-to-end system detecting time-series anomalies for simulated IoT data. Implemented a PyTorch temporal convolutional network, packaged with Docker, CI pipeline (GitHub Actions) and deployed to AWS Fargate behind an API gateway. Achieved 0.86 F1 on held-out synthetic anomalies; latency <120ms per request. Demonstrates data pipeline, model training, containerization, and serving.
2. PR to an open-source preprocessing library: Added a missing scalable feature-scaling transformer to a popular ML utilities repo (unit tests, type hints, and documentation). The contribution was accepted and shows collaboration, testing, and library-quality code.

This year allowed me to sharpen both research awareness and production skills; the artifacts are in my portfolio and reflect direct applicability to ML engineering responsibilities (modeling, deployment, monitoring, and reproducible code).