Open to opportunities

Leon Lampret

@leonlampret

Message

I build machine learning systems for forecasting, anomaly detection, and data-driven optimization.

Slovenia

Message

What I'm looking for

I'm looking for work combining programming and mathematics, where I can tackle hard computations, model and visualize data, develop efficient methods, learn and teach, and share ideas.

I've delivered machine learning projects now in use, including player profitability and churn forecasting, advertisement timing optimization, fraud-prevention anomaly detection, OCR, face recognition, and document metadata extraction.

I build these systems in Python using Redshift and PostgreSQL for data workflows, AWS SageMaker for training and inference, and RNNs, transformers, and gradient boosting for forecasting. I also communicate results through visualizations in Matplotlib, Plotly, Julia, and Tableau.

At the Jožef Stefan Institute's AI Lab, I worked on FactLog, predicting IoT sensor time series at an oil refinery to help improve final-product quality. I used PyTorch feed-forward neural networks and temporal fusion transformers for multivariate sensor forecasting.

Previously, at Queen Mary University of London, the Turing Institute, and the Defence Science and Technology Laboratory, I helped devise anomaly-detection methods for suspicious maritime trajectories using statistical, topological, and algebraic techniques. My mathematics PhD and research background continue to shape how I approach hard computations, modelling, visualization, learning, teaching, and sharing ideas.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist and ML Engineer

Current

Self-employed

Oct 2021 - Present (4 years 10 months)

Finished several projects that are in use today: time-series forecasting (predicting player profitability and churn), advertisement timing optimization (modelling daily/weekly habits of player engagement), anomaly detection (fraud prevention), and others.

Used Python for ML, redshift and postgres SQL DB for data wrangling, storing the results, scheduling runs and triggering alerts, and AWS SageM

Python SQL AWS Sagemaker Transformers Gradient Boosting Matplotlib Plotly Julia Tableau Forecasting Anomaly Detection Data Engineering Data Visualization Facial Recognition

Machine Learning Engineer

Jožef Stefan Institute, AI Lab

Mar 2021 - Sep 2021 (6 months)

Worked at an industry project FactLog (www.factlog.eu/) for predicting the timeseries of IoT sensor values at an oil refinery plant to improve the quality of final products using deep learning.
I used ANNs=artificial neural networks from the Python library torch to predict future values from many past values of a time-series of several sensors (multiple targets). I used basic 'feed-forward' ANNs

PyTorch Python Matplotlib Plotly GitHub

Data scientist

Queen Mary University of London; Turing institute; Defence Science and

Mar 2020 - Feb 2021 (11 months)

Worked on the project Topological Analysis of Maritime Data (www.qmul.ac.uk/maths/research/geometry-and-analysis-group/research-grant-projects-and-collaborations/#). We devised and implemented a method to detect anomalies in sea traffic (illegal or suspicious trajectories of vessels).
I used classical statistical methods in Python sklearn, scipy, faiss, KDEpy (KDE, NNS, cluster analysis, regressio

Python Geo Plotting