Hi, I’m Tomasz Lasota—a scientist, data geek, and AI enthusiast with a unique blend of expertise in Molecular Biology and Computer Science. With a PhD in Molecular Biology and an MSc in Computer Science specializing in Data Analytics, I focus on bridging the gap between wet-lab experimentation and computational insight.
My career began in the lab developing PCR-based diagnostics and nucleic acid extraction technologies, but I quickly became fascinated by the stories hidden in biological data. That curiosity drew me into the world of Python, machine learning, and statistical modeling. Whether it’s building classification models for PCR amplification curves, detecting anomalies in microfluidic pressure profiles, or interpreting metabolomics datasets using dimensionality reduction and clustering, I enjoy applying ML to solve meaningful biological problems.
My MSc research used artificial neural networks to model human biomechanics, but these days I'm more interested in making sense of high-dimensional biological data—especially in diagnostics and omics applications. I’m passionate about turning noisy lab readouts into actionable insights through clean, reproducible analysis pipelines.
Outside of science, I enjoy tinkering with 3D printing, automation projects, and cloud-native tools to accelerate discovery. And none of this would be possible without my dog, Lilly, who ensures I stay balanced with regular walks and reminders to unplug.
Collected time-series motion data from smartphone IMUs during resistance training. Built a 1D Convolutional Neural Network (1D-CNN) to automatically classify types of weight training exercises based on movement patterns.
Applied logistic regression, random forests, and support vector machines to predict depression risk based on lifestyle, sleep, food habits, and occupational stress indicators. Tuned models using cross-validation and feature selection techniques.
Used Kaplan–Meier curves and Cox Proportional Hazards models to explore survival trends. Trained ML models (e.g., XGBoost, LightGBM) to predict survival probability from high-dimensional clinical and demographic data with censored outcomes.
Developed supervised ML classifiers to categorize real-time PCR amplification curves into strong, weak, or failed signals, enhancing diagnostic accuracy and supporting QC decisions.
Used time series anomaly detection (e.g., moving z-scores, isolation forests) to detect pressure anomalies in microfluidic cartridges, identifying leaks or obstructions in early-stage prototypes.
Applied dimensionality reduction (PCA, t-SNE), clustering (k-means, DBSCAN), and Elastic Net regularization to identify key features in high-dimensional metabolomics datasets. Used statistical testing and visualization to distinguish metabolic profiles across experimental groups.