Bohan Li Email & Phone Number

Pittsburgh, Pa, Us

Worked as a full-stack engineer, developing a web interface for Astronomers to visualize Space missions and easily access/analyze/share data. Responsibilities:- UI design using Vue.js 2/3- Maintaining deployment stack using AWS CloudFormation- Designing database schema for Postgres SQL databases- Developing and maintaining Typescript backend interfaces with client and database

Madison, Wisconsin

I worked as a full-stack software developer at Epic, writing enhancement features and fixing issues for a software interface supporting pharmacist workflows related to home infusion, gaining experience in React and Typescript-based front-end frameworks and C#/MUMPS backend frameworks.

I implemented the command and control software for an HD camera flying aboard a CubeSat. My software handles UDP ethernet communication to the camera, pulling static image and video stream and writing data to memory registers for downlink. I utilized parallelization using multithreading to simultaneously pull images from the camera and write to registers. In addition, I oversaw the process of making my software open-source.I developed an application of spectral inference learning to improve imaging of exoplanets orbiting distant stars, where I utilized an algorithm to make use of Slow Feature Analysis to improve existing methods of PSF subtraction techniques to extract faint signals of exoplanets from image stacks of stars.

I worked on using GPU and CPU tools available on an HPC cluster to speed up two open-source machine learning libraries written in Python, Bingo and SMCPy. Throughout this project, I:- implemented GPU acceleration using custom kernels and CuPy library functionality - managed GPU calls to maximize computation on the GPU and minimize data transfer between devices- utilized GPU and CPU profiling tools (cProfile, nvprof) to monitor GPU usage and computation diagrams to identify performance bottlenecks- designed software to ensure backward compatibility with old library software, which did not depend on any GPU libraries

I worked as a software development intern for the Amazon Brand Registry, providing a mechanism of pulling registered trademark images from the United States Patent and Trademark Office (USPTO) regularly.Roles:- Designed full image ingestion process including back-end software to download and store images, as well as the front-end API for viewing pulled images- Implemented parallel bulk download of images utilizing Amazon Web Services (AWS) EMR through a Spark application- Managed security of the application by setting up a virtual private cloud to limit access to the EMR cluster while enabling the cluster to call services through the internet- Designed back-end data storage structure utilizing AWS S3 and Amazon internal Media Services to store images and downloaded USPTO zip files- Extracted trademark metadata from USPTO zip files, associating images with specific trademarks- Designed front-end API using Elasticsearch, allowing consumers to quickly query for images using trademark information

Tucson, Arizona Area

I worked on two projects relating to researching mathematical modeling of blood flow in tissue systems. The first project involved optimizing an algorithm to estimate blood flow in tissue networks. I developed a optimization technique utilizing sparse linear algebra techniques to produce a specialized iterative solver, significantly reducing the overall runtime of the algorithm for large datasets. A 97% improvement in computation time was seen for networks in the order of thousands of blood vessel segments. My second project was to develop a preprocessing algorithm for a 3D image stack of tissue, prior to segmentation of the image into a blood vessel network data structure. I developed an image smoothing algorithm based on 3D local percentile filtering. Each pixel samples nearby neighbors and undergoes a smoothing function based on a user-defined percentile threshold. I implemented this using multithreading GPU techniques using CUDA in C. I experimented with various sampling schemas and thresholding techniques and qualitatively evaluated effects of preprocessing on results of segmentation.

United States

I developed a customer-oriented new feature for AWS CloudFormation, designing the frontend API structure as well as the backend implementation. I created the feature through developing in several areas in CloudFormation's complex structure, incorporating AWS Lambda functions and AWS DynamoDB tables. I designed the DynamoDB table structure to best serve customers while minimizing costs to CloudFormation, and demonstrated potential uses for my implemented feature.

Master'S Degree, Computer Science

Bachelor Of Science - Bs, Mathematics And Computer Science