Teaching
Cloud for AI
Filotas's teaching combines lecture-based instruction in MLOps and applied machine learning with direct supervision of end-to-end student and thesis work. In Cloud for AI, he teaches the lifecycle of production-oriented ML systems while providing technical supervision across scoping, modeling, evaluation, and deployment planning. Students are expected to translate open-ended business problems into credible ML products, with attention to feasibility, system design, and real engineering trade-offs.
In addition to course supervision, he also guides master thesis work focused on end-to-end ML products and stable deployment architectures for applied research settings, including platform transition and support considerations.
Supervised End-to-End Student Projects
Selected individual projects from the course are shown below. These examples demonstrate technical supervision across scoping, modeling, evaluation, and deployment planning.
Adaptive Math Practice for Primary School Students (click for details)
Student build: Scoped an adaptive-learning ML system using student exercise data, trained and compared classifiers, designed rules for dynamic difficulty adjustment, and specified a FastAPI endpoint, Docker container, and Cloud Run deployment.
My supervision: Guided project scoping, reviewed model selection and evaluation choices, and assessed deployment trade-offs around latency and cost.
Outcome: Strongest implementation evidence in the set, including a referenced GitHub repository and a containerized prototype with reported low-latency inference.
Predicting Team Productivity in GitHub Projects (click for details)
Student build: Designed an end-to-end workflow using GitHub activity data to classify project-team productivity and flag low-productivity trends for managers.
My supervision: Supervised problem definition, feature engineering strategy, and evaluation choices such as Accuracy and F1-score.
Outcome: Educational end-to-end prototype with management-oriented benchmarking, best presented as supervised development rather than confirmed production deployment.
LLM-Based Support Case Review Assistant (click for details)
Student build: Designed a rubric-based QA assistant that ingests support case logs, generates structured JSON scorecards, and keeps a human reviewer in the loop before finalization.
My supervision: Reviewed feasibility, prompt design, risk controls, and human-in-the-loop integration for applied GenAI operations.
Outcome: Pilot-oriented prototype design with a measurable reviewer-time reduction target, presented as deployment planning rather than live deployment.
Bird Identification and Information App (click for details)
Student build: Proposed a computer-vision application where users upload bird photos, a cloud-hosted model predicts species, and a backend returns structured species information.
My supervision: Guided business framing, feasibility assessment, API-versus-custom-model decisions, and latency versus accuracy trade-offs.
Outcome: Strong cloud-application design with production-oriented systems thinking.
Music Recommendation System for Streaming Platforms (click for details)
Student build: Framed a recommender system using clustering methods to group songs and generate personalized recommendations tied to retention and churn goals.
My supervision: Supervised business problem scoping, evaluation metrics, and ethical reflection around personalization and user wellbeing.
Outcome: Applied recommender-systems project with clear business and ML evaluation framing.
EV Charging Station Demand and Placement Prediction (click for details)
Student build: Scoped an ML system to predict EV charging behavior and support charging-station placement through both a consumer-facing app and business dashboard.
My supervision: Guided the conversion of a societal and business problem into an ML product, including interface design, deployment thinking, and metric trade-offs.
Outcome: Dual-interface product design and deployment planning for a realistic infrastructure use case.
Course Content & Lectures:
The lecture material covers the practical MLOps lifecycle, from project scoping and data strategy to modeling, deployment, serving constraints, and monitoring. The emphasis is on engineering judgment: building strong baselines, choosing useful business and ML metrics, and making realistic production decisions.