Genomic data mining, machine learning, and image processing, in Python, C++ and Java.

Machine learning and image processing for MRI images using Keras, TensorFlow, VTK and ImageJ.

Built predictive models for phenotypes.

Put together an RNA pipeline from existing tools and some custom code for gene expression, allele specific expression, gene fusions and skipped exons in tumor tissue.

Built tools around scikit for convenient model training and validation.

Contributions to core and peripheral areas of search.

I taught the course Computational Biology with Java, and recently published the textbook Mastering Java through Biology. See http://www.petergarst.com, or click the link in the contact information section.

I developed the MoboMath system for recognition and editing of handwritten math expressions.

See www.petergarst.com/math.html for a link to a demo and a more extensive description, or click the link in the contact section.

It was based on a highly optimized Earley parser for stochastic grammars, and used neural networks, SVMs, and normal mixture models.

I made many upgrades to neural net software, including adding boosting and tanprop.

Editing operations were supported by the same architecture.

I received a patent for computational linguistics of mathematical expressions.

Projects included cursive handwriting recognition; Chinese handwriting recognition; and signature verification.

These were HMM based systems with computational linguistic input.

Also developed a distributed processing system to run training and testing jobs across all the workstations in the group.

Built a symbolic debugger for Yacc compatible grammars and parsers.

This included a compile time debugger to analyze grammars, and a run time debugger to trace and set breakpoints at significant points in the parse.

Customers included UC Berkeley, IBM, and others.