Profile
Research Scientist and Machine Learning Engineer focused on time series analysis, streaming data and machine learning systems. Experience designing and implementing end-to-end ML pipelines, from data ingestion and preprocessing to model training, evaluation and deployment. Background in information-theoretic analysis of predictability, deep learning for time series, and real-time anomaly detection with cloud-based streaming architectures.
Education
Ph.D. in Computer Science (Magna Cum Laude)
Feb 2021 – Aug 2025
University of Zurich, Switzerland
- Dissertation: “A Model-Agnostic Upper Bound for Univariate Time Series Prediction”
- Advisors: Prof. Michael H. Böhlen, Prof. Sven Helmer
- Focus on predictability quantification, information-theoretic bounds, deep learning for time series, transformer architectures and scalable ML systems.
M.Sc. in Computer Science (110/110)
Sept 2018 – Sept 2020
Free University of Bolzano, Italy
- Thesis: “Advanced Deep Learning Approaches with Transfer Learning for Industrial Time Series Forecasting in Predictive Maintenance”.
- Capstone project: cloud-based threshold anomaly detection system using Apache Kafka and Apache Flink (Java/Scala) on streaming sensor data.
B.Sc. in Computer Engineering
Sept 2014 – Mar 2018
Politecnico di Torino, Italy
PhD Research – Selected Impact
- Developed a predictability upper bound for time series forecasting tasks to understand when additional model complexity is likely to bring limited benefit and when there is room for improvement.
- Used the bound to distinguish between intrinsic data noise and model limitations, indicating whether effort should focus on data collection and feature engineering or on model design and training.
Professional Experience
Doctoral Researcher – Machine Learning & Data Science
Feb 2021 – Aug 2025
University of Zurich, Switzerland
- Designed and implemented end-to-end ML pipelines for time series analysis, from data ingestion and feature preparation to model training and evaluation.
- Worked with heterogeneous data (sensor streams, static profiles, temporal measurements) in a common framework for systematic benchmarking.
- Set up CI/CD elements for ML, including automated tests and training/evaluation jobs, to improve reproducibility.
- Implemented deep learning models (LSTM, N-BEATS, N-HITS, PatchTST, Transformers) for forecasting and anomaly detection tasks.
- Developed an open-source Python toolkit for predictability estimation, model benchmarking and forecasting evaluation.
- Used distributed computing and multi-GPU training (CUDA, PyTorch) to run experiments on larger datasets.
Data Engineering & ML System Deployment – Predictive Maintenance
Jun 2019 – Sept 2020
Industrial Predictive Maintenance Service – Durst Phototechnik
- Helped design and deploy a predictive maintenance pipeline for more than 50 industrial sensors in an on-premises environment.
- Set up time series data storage using InfluxDB and TimescaleDB with schemas suited for high-frequency sensor data.
- Implemented real-time data pipelines and anomaly detection algorithms to monitor equipment conditions.
- Built ETL processes combining sensor data with maintenance logs and operational parameters.
- Worked with operations and maintenance teams at Durst Phototechnik to understand requirements, review dashboards and adjust alert thresholds.
Cloud & Streaming Data – Anomaly Detection (Capstone)
2019 – 2020
M.Sc. Capstone Project
- Implemented a cloud-based anomaly detection system for streaming sensor data using Apache Kafka and Apache Flink (Java, Scala).
- Developed threshold-based anomaly detection rules and real-time alerting.
- Deployed the streaming pipeline in a cloud environment with producers, consumers and fault-tolerant processing.
Database Administration & Systems Design
Sept 2021 – Aug 2024
University of Zurich – Database Systems Infrastructure
- Administered a PostgreSQL database used by more than 300 concurrent users in a teaching environment.
- Set up a multi-tenant authentication system with role-based access control.
- Designed schemas and indexing strategies and improved performance of key queries.
Technical Skills
Data Science & ML
- Time series forecasting and anomaly detection
- Data wrangling and exploratory analysis
- Deep learning: PyTorch, TensorFlow
- Models: Transformers, LSTMs, N-BEATS, N-HITS, PatchTST
MLOps & Engineering
- CI/CD for ML (Jenkins, GitLab CI)
- Docker, experiment tracking
- Git, GitHub; Linux/Unix
- Distributed and multi-GPU training (CUDA)
Data Engineering & Databases
- ETL/ELT pipeline design
- Apache Kafka, Apache Flink (Java/Scala)
- InfluxDB, TimescaleDB
- PostgreSQL administration and optimization
Programming & Tools
- Python (PyTorch, NumPy, Pandas, Scikit-learn)
- Java, Scala, SQL, Bash, C, MATLAB
- Matplotlib, Seaborn, Plotly; Jupyter, VS Code, LaTeX
Publications
- J. Mohammed, M. H. Böhlen, S. Helmer (2024). “Quantifying and Estimating the Predictability Upper Bound of Univariate Numeric Time Series.” Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (SIGKDD 2024).
- J. Mohammed (2025). “Multi-Horizon Predictability Upper Bound Time Series Forecasting.” In preparation, submitted to ICML 2026.
- J. Mohammed, M. H. Böhlen, S. Helmer (2025). “Efficient Subsequence Matching in Time Series for Entropy Rate Estimation.” In preparation for IEEE TKDE.
Teaching & Mentorship
- Teaching Assistant, Database Systems, University of Zurich (Sept 2021 – Aug 2024): tutorials for 150+ students on SQL, database design and query optimization; preparation of assignments and exam material.
- Supervised 3 M.Sc. and 2 B.Sc. theses on machine learning for sequential data and information-theoretic approaches to time series forecasting.
Projects, Awards & Other
- Time Series Predictability Toolkit: open-source Python library for predictability bounds, model benchmarking and forecasting evaluation.
- Ph.D. degree awarded with Magna Cum Laude distinction, University of Zurich (2025).
- M.Sc. scholarship for semester exchange program (2019).