Jun Dai Email & Phone Number

Texas, United States

Somerville, Massachusetts, Us

• Target identification and validation using in-house causal machine learning platform• Develop deep learning models with cancer dependency map and connectivity map data for target validation and biomarker discoveries• Using multi-omics data from public domain and data vendors for oncology studies• Using single cell RNA-seq data for target identification at both cell and pseudo-bulk level

Frisco, Texas, Us

• Solid tumor metastasis prediction based on mutation, fusion and copy number alternations;• Single strand DNA aptamers design targeting the neutralization of viral or allergen proteins;• Proteomics data enrichment and pathway analysis, etc.

Lincoln, Ne, Us

• Lead 3 collaborative projects on simulation side with 4 teams nationwide, published 6 peer reviewed papers, filed one patent application.• Provided a solution to directly predict the efficiency of 2D carbon-based materials from structures through a combination of machine learning and high-throughput calculations using Python with R-squared score over 95% on a small dataset with about 200 samples. Transfer learning was used to treat the small dataset.• Collected a dataset of about 2500 samples for 2D perovskites by combining in-house data and running SQL queries from open-sourced material databases. Trained and optimized deep learning graph convolutional neural network models to predict the material properties with R-squared score over 90%.

Lincoln, Ne, Us

Teach general chemistry for a class of 185 students.

Lincoln, Ne, Us

• Using Python, Request and Beautiful Soup, scraped about 4000 house information in the Omaha, Lincoln and Des Moines area from zillow. Complemented the nearby venues data using Foursquare API. Performed both regression and classification using scikit-learn, the best R-squared for regression is 89%, and the accuracy for classification is about 86%.• Using Python scikit-learn and pySurvival, trained random forest and xgboost models to predict the patient mortality based on the queried data, the best AUROC achieved is 0.87; performed survival analysis to predict the patient death after admission to ICU over a 40 days' period, the best model has a mean absolute error of about 1 patient. Built a MIMIC-III database using postgresSQL on a local server for clinical data management.• Using VGG16, VGG19, MobileNet or Inception networks as based model, and pre-trained parameters from ImageNet, performed transfer learning in the framework of tensorflow to diagnosis pneumonia disease in 5863 chest X-Ray images. PCA was also applied to accelerate the training, best recall achieved was 98%, together with a F1 score of 94%.• Implemented an interactive financial management system using Java and SQL. The system has functions of modifying, retrieving user and assets data, and providing financial reports for portfolios, such as total value, return rate, aggregate risk, etc.

San Francisco, Ca, Us

• Delivered a data process strategy to a client company for feature engineering and product categorization.• Using Python, NLP sentence embedding based on Google's Universal Sentence Encoder and tensorflow, designed a model to classify products into a taxonomy customized for a consulting client.• Categorized 151256 products from 9 different retailers into 708 customized categories, significantly increased the taxonomy granularity from 2-tier to 5-tier.

•Taught two sections of general physics for a class of 50 undergraduate and graduate students with diverse background.•Designed and implemented new course curriculum, lesson plans, and assessment tools for diverse group of students.

Lincoln, Ne, Us

•Using particle swarm optimization algorithm for genetic evolution and random sampling, systematically studied the 2D structures of boron and boron-related materials, the ground state structures were obtained by analyzing thousands of the generated structures. The study inspired the synthesis of Borophene, and the 2D boron work was cited by Wikipedia.•Theoretically predicted the high- and anisotropic carrier mobility in 2D titanium trisulfide for the first time, developed and led collaboration with experimental groups for the synthesis and characterization.•Computationally studied phosphorene as a donor material in optoelectronic applications for the first time, this work is cited by Wikipedia.

Hefei, Anhui, Cn

•Developing an in-house software Order-N Quantum Chemistry Package for Large Scale Ab Initio Simulation (ONPAS) for first-principles calculations.•Predicted a new 3D boron phase, evaluated its stability and relative formation entropy as compared to other 3D boron phases.•Using lattice dynamics and electron-phonon coupling techniques, predicted the existence of high temperature (above liquid nitrogen) superconductivity in 2D B2C. •In charge of the administration of in-house computing clusters.

Doctor Of Philosophy (Phd), Chemistry

Bachelor'S Degree, Physical Chemistry