Hi there! My name is Jin

I'm a student passionate about machine learning solving impactful problems developing tools for everyday people

  • Courses
  • Research
  • My Skills

About Me

Welcome! I'm a math and CS major at UChicago ('22). I'm passionate about applying machine learning (ML) to solve impactful problems, and I have over two years of experience building ML products.

Currently, I'm a deep learning research assistant at TTIC, where I work with professor Jinbo Xu to use deep learning to predict protein structures. I'm also a ML research assistant at UChicago Medicine, where I work with professor Ishanu Chattopadhyay to build ML models to solve biological problems like improving vaccine targets, predicting cardiac arrests, and quantifying pandemic risks. I'm the president and founder of ML @ UChicago, a community based around democratizing ML knowledge.

In my spare time, I like to go outside and play soccer, go on hikes, and camp in forests!

Skills

Python

95%

TensorFlow

90%

Scikit-Learn

95%

Pytorch

60%

AWS

60%

GCP

50%

Numpy

95%

Pandas

80%

C

70%

Git

60%

R

40%

Courses Taken

Before beginning UChicago, I recieved 1000 credits from high school courses and examinations (100 credits is equivalent to 1 course).

I've taken many undergraduate CS courses like discrete math, theory of algorithms, formal languages, complexity theory, computer systems, computer architecture, and parallel computing. I've also taken (or currently taking) graduate level CS courses like machine learning, deep learning, deep learning systems, advanced NLP, computational imaging, multivariate data analysis, and quantum computing.

I skipped 7 math courses in UChicago and completed the selective year-long honors real analysis sequence in my first year.

I've also supplemented my college degree with online courses from Coursera. I completed courses for topics including AWS, GCP, and leadership.



Research Papers

Papers Under Peer Review:

Preparing For the Next Pandemic: Learning Wild Mutational Patterns At Scale For Analyzing Sequence Divergence In Novel Pathogens by Jin Li, Timmy Li, Aaron Esser-Kahn, Ishanu Chattopadhyay, 2020. PDF

Predicting Appropriate ICD Therapies with Daily Remote Home Monitoring in the IMPACT Trial: A comparison of classical versus machine-learning based approaches by Curtis Ginder, Jin Li, Roderick Tung, Ishanu Chattopadhyay, Gaurav A. Upadhyay, 2020

Development and Validation of Computerized Adaptive Assessment Tools for the Measurement of Posttraumatic Stress Disorder by Lisa A. Brenner, Lisa M. Betthauser, Molly Penzenik, Anne Germain, Jin Li, Ishanu Chattopadhyay, Ellen Frank, David J. Kupfer, Robert D. Gibbons, 2020

Improved Protein Structure Prediction by Deep Learning Irrespective of Co-evolution Information by Jinbo Xu, Matthew Mcpartlon, Jin Li, 2020. PDF

Study of Real-Valued Distance Prediction For Protein Structure Prediction with Deep Learning by Jin Li, Jinbo Xu, 2020

Arxiv Preprints:

Universal Transforming Geometric Network by Jin Li, 2019. PDF

My Portfolio

Here are some of the projects that I have done (or am currently working on). I open sourced most of them, but some are privated until a research paper has been published or the project is finished.

  • All Works
  • Structural Biology
  • Healthcare
  • Other
Universal Transforming Geometric Network Architecture

Structural BiologyUniversal Transforming Geometric Network

I developed a novel deep learning method to help predict a protein's structure. This improved an existing method by 5% while cutting training time by 56%. It is also seven orders of magnitude faster than traditional protein prediction methods.

Quasinet

Structural BiologyQuasinet For Influenza and COVID-19

With my professor, I developed the Quasinet, a novel unsupervised ML algorithm for biological sequences. In the research paper where I was the lead-author, we showed that this model can effectively investigate the origins of COVID-19 and improve the World Health Organization's vaccine target for Influenza by up to 81%.

Quasinet for Pandemic Risk

HealthcareQuasinet for Pandemic Risk

I showed that the Quasinet can correctly predict the 2009 Influenza outbreak and that the Quasinet can quantify the relative risk of animal-human transmission of Influenza, Ebola, HIV, and SARS-COV-2. Another research paper where I am the lead author is going to be released soon.

Infant Microbiome

HealthcareQuasinet For Infant Microbiome

I'm investigating how the Quasinet can be used to find relationships between gut measurements in an infant and how that infant grows in the future.

CASP 14

Structural BiologyCASP14 Competition

I rewrote a major portion of the code base to use TensorFlow 2.0 in Xu's lab for protein folding to improve performance while reducing training time by 45%. The models that I trained are being used to predict and fold proteins for the CASP14 competition. I will co-author a paper when the competition is over.

CASP 14

Structural BiologyRaptorX vs DeepMind's AlphaFold on Free Modeling

We beat AlphaFold's contact accuracy by raising the contact F1 score from 41.9% to 52.1%, or a 24% improvement. We also beat their average TM Score by a significant amount (0.640 vs 0.583).

CASP 14

Structural BiologyImproved Protein Structure Prediction W/ and W/O Co-Evolutionary Data

We showed that our protein structure prediction pipeline outperforms existing methods with and without using co-evolutionary data. I'm a co-author to this paper that is under review.

CASP 14

Structural BiologyComparison of Real Valued and Discrete Inter-residual Distances For Protein Structure Prediction

We compare the pros and cons of real vs discrete values for protein structure prediction. I'm the lead author to this paper that is under review.

CASP 14

OtherMeasuring Cognitive Dissonance

We showed that the quasinet can effectively measure cognitive dissonance among a population and can be used to simulate how beliefs change over time.

Infant Microbiome

Structural BiologyGAN Sampling For Protein Folding

I'm developing a novel GAN network that can effectively sample multiple predictions to improve protein folding.

CAT PTSD

HealthcarePost-traumatic Stress Disorder Diagnosis

I deployed a trained machine learning model onto a web server to improve efficiency of PTSD diagnosis. I'm co-authoring a paper on PTSD diagnosis that is currently under review.

CAT PTSD

HealthcareCardiac Shock Prediction

I trained a deep neural network that improves cardiac prediction from 0.71 AUC to 0.93 AUC. I'm a co-author to this paper, which is currently under review.

Contact Map Prediction

Structural BiologyDeep Contact Prediction

I replicated an existing research paper that uses deep residual convolutional neural networks to predict a protein contact map.

HealthcareWater Pump Prediction

I built visualization tools, performed feature engineering, and built machine learning models like random forests, gradient boosting, and decision trees to predict which water pump is not functional. My model is within 2% of the best available model.

Structural BiologyProtein 3SRP Mutation

I used molecular dynamics to see whether a certain mutation to RiVax has the potential to act as a vaccine for ricin, a deadly poison.

OtherQuantum Circuits

I wrote a library to simulate quantum circuits and compile circuits written in Python into machine code for quantum computers.

OtherSOTA Convolution Layers

I replicated various state of the art convolution layers in TensorFlow 2.0.

Chicago bean

Social SciencesArtistic Style Analysis

I built deep neural networks to measure styles, moods, and emotions of Chicago artworks.

Yelp

Social SciencesExtracting Culture From Yelp Data

I built A.I. language models to understand how Yelp data can provide insight to how popular culture is changing over time.

See more on Github.

Work & Education

It's only the beginning of my journey! But here are the major milestones so far.

Jun 2018

Graduated from Stuyvesant High School in NYC with honors.

Oct 2018

Started college at the University of Chicago

Oct 2018

research assistant at the Scenes Project

I built deep neural networks to measure styles, moods, and emotions of Chicago artworks. I also built A.I. language models to understand how Yelp data can provide insight to how popular culture is changing over time. End date: March 2019.

Feb 2019

Software Engineer Intern at Finch

I built a website with a backend for a trading dashboard. I also found ways to use machine learning and natural language processing to extract data from financial documents. End date: June 2019.

March 2019

Machine learning research assistant at ZED Group.

I'm involved with many ongoing machine learning projects, including developing the Quasinet, predicting cardiac arrests, predicting PTSD, using machine learning to improve vaccine targets for Influenza and COVID-19, and much more!

Jun 2019

Deep Learning Intern at ShanghaiTech University with Professor Jie Zheng

I developed a novel deep learning algorithm to predict protein structures. End date: August 2019.

Aug 2019

Deep Learning research assistant at TTIC with professor Jinbo Xu

I help rewrite legacy code to speed up deep learning performance. I'm also involved in building novel deep learning algorithms to predict protein structures.

Jun 2020

Founder and President of the Machine Learning Community

I started the machine learning community to teach machine learning to everyone, create an open community for learning, and build meaningful projects to solve important problems.

My Motivational Quotes

The things that make me different are the things that make me.

Winnie the Pooh

In the book of life, the answers aren’t in the back.

Charlie Brown

Just keep swimming.

Dory from Finding Nemo

Contact Me

If any of my work sounds interesting to you, please do contact me! I would be excited to work with you! You can reach at: jinli11@uchicago.edu