Cristian Ghitu

Planned sabbatical, dedicated to travel and preparing for a transition to independent consulting. Period used to design an online introductory Scala course and update my ML Engineering and GenAI technical stack.

- Online Course Creation: Design and development of an online introductory Scala course (videos, presentation materials, interactive exercises, and quizzes). Writing automated unit tests in Scala to evaluate student solutions and using the host platform interface for deploying self-assessed quizzes.

- Upskilling: ML Engineering and GenAI.

MLE: containerization (Docker), workflow orchestration (Apache Airflow), model/API serving (FastAPI), CI/CD (GitHub Actions).

GenAI: LLM, pytorch.

- Certification: Obtained AWS Certified Cloud Practitioner certification.

Carried out Data Science projects in Agile Scrum, from scoping to industrialization.

Key Projects:

- Risk Estimation: Creation of industrialized data and prediction pipelines for risk score calculation.

Data cleaning, preparation, and feature engineering with PySpark and model training with scikit-learn and other ML libraries, using SHAP to explain predictions.

Structuring processes into pipelines using Kedro, with Great Expectations for quality control.

Using MLflow for model versioning and performance tracking.

- Transaction Automation: Creation of an industrialized model to reduce the time advisors spend on transaction processing.

Transitioning exploratory Python code to industrialized Scala code.

- Regulatory KPIs Calculation: Processing massive data with PySpark for the calculation of regulatory indicators.

- Billing Anomaly Detection: Application of unsupervised models for anomaly detection on billing lines.

- IT Incident Classification (PoC): Application of NLP techniques to group and classify IT incidents by theme.

Using Spark and Scala with OpenNLP and StanfordNLP to distribute and reduce text preprocessing time.

Conventions and Sharing:

- Standardization: Implementation of standards for ML projects to accelerate the transition from PoCs (Jupyter notebooks) to industrialized applications, establishing conventions for exploratory code.

- Mentoring and Lessons Learned: Organizing meetups and sharing lessons learned as a member of developer communities.

Mentoring junior team members during their onboarding.

Technical Stack: Python, Scala, Apache Spark (PySpark), Hadoop, Kedro, MLflow, Great Expectations, scikit-learn, Hyperopt, XGBoost, H2O, SHAP, Git, Dataiku.

Final Project: Predictive Maintenance.

- Exploration and application of anomaly detection algorithms on time series data from application logs to enable predictive maintenance of application servers.

Methods explored: double seasonality Taylor model, LSTMs, and autoencoders using Python and Keras.

Technical Stack: Python, Keras, scikit-learn, pandas.