research

ongoing

a primary focus of the gomez lab is the complex biological 'signal processing' network of proteins known as the Kinome. i am currently working with professor shawn gomez to investigate the construction, architecture, and applications of kinase interaction networks (kin) from a systems biology perspective

One example in progress is a 'kinome api' with connections to phenotype regression exploring the architecture of the kinome spawning from work of recent unc graduate kyla a.l. collins

Sample Kinase Interaction Network — visualization of a(n outdated) kinase interaction network with 473 nodes (protein kinases) and ~3600 edges (protein-protein interactions) using the forceatlas2 algorithm in javascript library sigma.js. sizes are logarithmic of degree count for each node; colors represent candidate groupings

additional areas of ongoing research include applications of artificial intelligence in both single-cell rna-sequencing data and medical image analysis.

Differentiation UMAP — umap dimensionality reduction time lapse of mouse embryonic fibroblasts (mefs) differentiating into a variety of other cell types. [induced pluriplotent] stem cells (ipscs, in yellow) are a 'target' to diffferentiate into, although many mefs specialize into other cell types. data visualized are a subset of data collected by schiebinger et al (2019)

past

one of my first big self-driven data science projects was a capstone for Udacity's machine learning [engineer] nanodegree (mlnd). i decided to explore the machine-generated clusters of french geocode demographics (a geocode similar to zipcode in the usa) to official government-created classifications of 'urban' and 'rural'

French Geocode Clustering — outputs of different scikit-learn clustering algorithms on multiple data preprocessing pipelines and cluster counts.

presentations and whatnot

colorful powerpoints spark joy in me. feel free to download a few of mine (and some other things, too) ~

from comp790-158 spring 2020

current special topics course in nlp at unc with professor snigdha chaturvedi

hour presentation on chambers and jurafsky (2008)

from ece590 fall 2019

a fantastic hardware acceleration course i took at duke university with professor lisa wu wills

on roofline, a powerful bottleneck analysis tool for scientific computing
on gables, a roofline model for mobile systems on chips
on the tpu (v1), the custom asic from google for machine learning
on darwin, a long-read genomics accelerator
on mondrian, a relational database accelerator

from bmme890 fall 2019

an innovative biomolecular engineering course at unc with professor koji sode

on protein language models, a promising deep learning architecture, and use cases with erbb2 and pertuzumab

from comp776 fall 2018

a fascinating computer vision course at unc with professor alex c. berg

a very entertaining course report on an instance segmentation project PotatoNet (yes, i manually labeled over 1000 potatoes)