Russell Ro
#
hirsqrd@gmail.com
hirsqrd.com
ï
LinkedIn
§
Github
Education
University of California, Berkeley / University of California, San Francisco 2022 - Present
Ph.D. in Bioengineering 3.6 GPA
•
Relevant Coursework: Convex Optimization, Principles & Techniques of Data Science, Intro to Computational Biology
•
Fully funded by NSF Graduate Research Fellowship Program (GRFP).
University of California, San Diego 2018 - 2022
B.S. in Bioengineering 3.9 GPA
•
Relevant Coursework: Semiconductor Physics, Systems & Control, Modeling & Computation, Numerical Methods, Statistics
•
Full-ride scholarship from Jacobs School Scholarship. Approximately 10 recipients every year.
Experience
University of California, San Francisco August 2023 – Present
Graduate Student Researcher San Francisco
•
Exposed clinical large language model (LLM) vulnerabilities by designing a novel distance weighted embedding sampling
algorithm (standard: identifies ∼ 10% vulnerabilities; our method: ∼ 17%). Developing new prompt engineering based
algorithms to overcome these limitations to ensure safer, more robust LLMs.
•
Identified and ranked
feature importance of neural correlates that are most explanatory of motor behavior by interpreting
logistic regression models trained to decode keypresses from 236 electrocorticography (ECoG) signals.
•
Currently conducting independent research project investigating meta-learning, factor analysis, and regime shift
identification methods for interpreting weight changes and comparing models trained on similar tasks.
Novartis Institutes for BioMedical Research October 2020 – July 2022
R&D Engineering Intern San Diego
•
Proposed and spearheaded project to scale up automated cardiovascular drug screening and integrate system into Novartis’
high-throughput screening (HTS) pipeline (previous system: 8-channels; my system: 96-channels; 12x improvement).
•
Collaborated with Novartis Data Science team to integrate the project, thereby enhancing their computer vision-based drug
discovery platform.
•
Designed electrical stimulation control system and PCB (previous system–$2000; my system–$100; 20x cost reduction) which
improved frequency response for signal route switching (previous system–response on the order of seconds; my system–microseconds
i.e. 10
6
x speed improvement).
Instrumentation Laboratory, A Werfen Company June 2020 – September 2020
R&D Engineering Intern San Diego
•
Influenced design changes of next-gen hemostatic diagnostic devices by designing a thermal testing system which aided in the
discovery of a key engineering design flaw that has led to damaging blood samples.
•
Discovered that devices overheat blood samples (standard: 37°C; device: 42°C; ∼ 10% excess) by conducting thermal
experiments and analyzing time series data and frequency responses from multiple devices.
•
Derived interpretable metrics from noisy data by proposing heat transfer model used to fit to data and generate a sample
distribution of model coefficients.
Projects
MitoGAN
•
To capture a biophysical model of mitochondrial dynamics in silico, processed and trained a generative model on image data of
mitochondria microscopy images.
•
Built an image processing pipeline and trained a Generative Adversarial Network (GAN) to capture the distribution of input
features within a model, generate synthetic training data from microscopy images (original dataset: ∼ 600 images; generated:
∼ 30000; 50x generated), and interpret the most salient geometric features for identifying mitochondria.
CAR T-cell Protein Sequence Discovery
•
Scaled compute for DNA sequence alignment using AWS EC2 instances (before: 1x local machine; after: 4x cloud machines)
for identifying novel immunoreceptor (Chimeric Antigen Receptor, CAR) domains/motifs.
Anime Recommendation Using Collaborative Filtering
•
Generated user anime recommendations by predicting missing scores within a sparse matrix (137M predictions; filled 98.9% of
matrix) using user-based collaborative filtering (cosine similarity) on MyAnimeList user rating data.
Publications
Zero-shot sampling of adversarial entities in biomedical question answering Feb 2024
ScanAlert: Electronic Medication Monitor and Reminder to Improve Medical Adherence Jan 2019
Skills
Languages: Python, SQL, C++,
MATLAB, LabVIEW
Data Analysis: NumPy, Pandas, Excel
Data Visualization: Matplotlib, Seaborn
Machine Learning: Scikit-learn, PyTorch
MLOps: TensorBoard, MLflow
Version Control: Git, GitHub, Data
Version Control (DVC), DagsHub
Cloud: AWS (EC2 & S3)
CAD: Altium, Eagle, Cadence PSpice,
SolidWorks, Inventor, Fusion 360, Creo
Prototyping/Mnfg.: 3D Printing,
Arduino, Photolithography
Data Science: Exploratory Data Analysis
(EDA), Visualization, Data Cleaning,
Imputation, Data Augmentation,
Train-Test-Validation Split, Principal
Component Analysis (PCA), One Hot
Encoding (OHE), Factor Analysis
Deep Learning: Neural Networks,
Backpropagation, Supervised, Unsupervised,
Self-Supervised Learning, Recurrent Neural
Networks (RNNs), Residual Networks
(ResNets), Autoencoder, Transformer,
Pre-training, Fine-tuning, Transfer Learning,
Curriculum Learning
Time Series: Point Change Detection,
Regime Shift, Autoregression (AR, ARMA,
ARIMA), State Space Modeling
Signals Processing: Fourier Transform
(FFT), Hilbert Transform, Bandpass Filter
Natural Language Processing:
Word2Vec, Spam Filter (Naive Bayes),
Sentiment Classification, Question
Answering, Embedding, Transformer, Large
Language Models (LLMs)
Computer Vision: Convolutional Neural
Networks (CNNs), Image Classification,
Edge Detection, Segmentation
Recommendation Systems:
Collaborative Filtering (Cosine Similarity,
Singular Value Decomposition)
Honors and Awards
NSF Graduate Research Fellowship Program (GRFP) May 2022
National Science Foundation
Tau Beta Pi May 2020
Tau Beta Pi Engineering Society
Jacobs School Scholar Sep 2018
Joan and Irwin Jacobs Foundation
•
Full-ride scholarship to attend UCSD Jacobs School of Engineering. Approximately 10 recipients every year.
UNITE Research Scholarship June 2017
Army Educational Outreach Program
•
Fully-funded 6-week research program at the University of Nevada, Las Vegas.
International Science & Engineering Fair (ISEF) Finalist May 2017
Intel
•
Selected to represent Nevada at the largest international pre-college scientific research conference.
