Alex R. Sathler

Models & Methods: Dermis & Epidermis • Predictive diagnostics • FLAME microscopy • Content-aware image restoration • Fluorescence lifetime imaging • Computer vision

Languages & Frameworks: Python • Tensorflow • UNet • ONNX • MLFlow • MLOps • OpenCV • WSL • Jupyter • Command-line

Impact:

• Developed a robust denoising computer vision model and associated command-line tool that reduced acquisition timelines by 75% in FLAME imaging of epidermal fluorescence for predictive skin diagnostics.
• Created de-novo ML pipelines for dataset curation, model training, model evaluation, model performance tracking, and model deployment on intranet-based server infrastructure.

Models & Methods: Mammalian bioreactors • Bioprocess engineering • Instance segmentation • Product development • Digital staining • Cell viability • quantitative phase imaging (QPI) • CHO cells • HEK293 cells • E. coli • S. cerevisiae • UX design

Languages & Frameworks: Python • C++ • Batch • Bash • Tensorflow • PyTorch • Detectron2 • TensorRT • ONNX • Vertex AI • Lightning AI • OpenCV • WSL • Qt

Impact:

• Trained a 90% accurate computer vision model enabling real-time live cell viability quantification in mammalian cell bioreactor, allowing real-time culture monitoring and informing feedstock delivery to reduce waste by 3-5% in pharmaceutical manufacturing pipelines.
• Integrated product development into venture capital pitches and customer outreach, driving investment opportunities in a $1.5 billion market while increasing marketing engagement by 70X and establishing 3 OEM partnerships.
• Reported on model performance in prediction of cell viability across three eukaryotic and two bacterial, driving receipt of SBIR funding for commercialization of QPI as a novel label-free imaging modality.

Models & Methods: AI/ML • Computer vision • Phase separation • Aging • Murine primary cortical culture • Murine DRG culture • Airyscan confocal microscopy • Live & fixed immunofluorescence • MTT assay • Extracellular flux analysis • High-performance computing • Plasmid amplification • HEK293 culture • R-Loops

Languages & Frameworks: Python • Tensorflow • Bash • HPC • Slurm • OpenCV • ImageJ / FIJI • Jython

Impact:

• Revealed redox-mediated separation of mitochondrial DNA into transcriptionally deficient phenotypes, suggesting a novel explanation for aging-associated neuronal energy crises and cognitive declines.
• Achieved 96% pixel-level classification accuracy and 86% intersection-over-union semantic segmentation ML model with a small dataset of six manually annotated images.
• Elucidated novel mechanistic hypothesis for aging-associated reductions in mitochondrial transcription and translation through literature review to reveal a novel experimental avenue for a senior NIH investigator.

Models & Methods: Glioblastoma multiforme (GBM) • U87 cell line • Western blotting • Crystal violet assay • Genetic code expansion • Redox biology • Protein engineering • Confocal microscopy • Python • Jupyter • OpenCV

Languages & Frameworks: Python • Jupyter • OpenCV • Matplotlib

Impact:

• Unveiled a metabolic switch in Glioblastoma Multiforme (GBM) caused by the post-translational nitration of specific tyrosine residues in heat shock protein 90 that mimic the Warburg effect by reducing oxygen consumption in tumor periphery and increasing glycolysis in the tumor core.
• Built a qualitative suite for protein distribution validation in any 3D culture model, enabling cost-effective visualization of spatio-temporal gene expression in cell biology, pathology, and tissue modeling.
• Established low-cost method to understand protein distrubution in three dimensions, reducing barriers to discovery given inacessibility of expensive software analysis suites.

Languages: Python • SQL • R

Frameworks: AWS • GCP • Jupyter • Pandas • RStudio • VCF

Domains: Bioinformatics • Oncology • Cloud Computing • Statistics • Visualization

Destription: Implemented genome fingerprinting and comparison into a custom tissue processing pipeline for patient-personalized CAR-T therapy in the nation's 11th largest health system. Integrated into existing AWS workflow as a quality measure to conclusively determine whether a tumor sample comes from a patient or not.

1. Created and fingerprints from the hundreds of tumor & normal tissue samples.
2. Assessed quality control efficacy with statistical and visualization methods.
3. Enabled integration into AWS pipeline for high-throughput quality control.