As a dedicated software engineer, I specialize in developing software solution in Python/Go to address optimization challenges in maritime and medical Imaging.
Experience
2024 — Now
United States
Identified a gap in a fast growing market with rapidly changing demand through in-depth research. Leveraging my previous experience to address the needs of a small but specific target audience, while acquiring new skills (including new programming language).
Design and developing services from scratch, including the python/go backend to kotlin client with fast iterations.
2023 — 2024
New York City Metropolitan Area
Maintained job search momentum during market downturn following company(a startup which I really enjoyed working at and stayed until the last second) closure.
Dealt with unexpected failure during interviews on leetcode-style questions that I usually solve without problem. Stay professional even when things are not working out during interview.
Maintained composure during technical challenges, even for soul-searching behavior questions. Clearly communicated thought processes to demonstrate problem-solving abilities.
2022 — 2023
New York City Metropolitan Area
● Developing a critical ETL (Extract, Transform, Load) service pipeline in Python and Go for weather forecast data, essential for core routing and speed optimization products.
● Developed and maintained services for ingesting weather data from NOAA, processing over 3,800 files (43GB) daily. Converted this data into a searchable tile format and stored it in AWS S3 buckets, accumulating a total of 110 TB across 17 million files.
● Navigating a complex multilayer caching system to integrate new data features and enhance runtime performance. Fine tuning the integration of weather service to our optimization product.
● Creating a data parser that outputs JSON files directly utilised by the frontend component of our service through AWS CloudFront, balancing speed and visualisation accuracy.
● Continuously improving the data-driven email alerting system using SendGrid, addressing inconsistencies between client-recorded emails and our metrics. Troubleshoot missing records related to authentication.
● Enhancing system robustness by implementing tracking redundancy and health monitoring through Datadog and Mezmo.
● Managed and debugged a critical Redis cache system, integral to key services. Served as the primary point of contact for resolving issues, including feature outages and performance degradation. The cache service handled approximately half of the total data, encompassing 32 million key-value pairs and processing 12.5 billion sensor data entries per year, contributing to a total of 50 billion data points.
● Translating industrial standard emission calculation documents into code and continuing to develop emission calculations according to IMO (International Maritime Organization) standards (Python and Go).
New York City Metropolitan Area
Software Development for Machine Learning MR Image Reconstruction
Developing simulation pipeline to generate realistic breast perfusion images. Developing multi-coil array, non-cartesian MRI raw data simulation pipeline for accelerated simulation on GPUs with PyTorch.
Reconstruction Software Development: Developing and implementing iterative reconstruction model with convolutional neural network as regularizers. Using simulated data to train the model with GPUs on computer clusters to reconstruct breast perfusion images.
Exploring alternative architecture (GRU，LSTM, U-Net etc) and loss function optimization to increase the reconstruction accuracy of reconstruction and its robustness.
fastMRI (In collaboration with Facebook AI)(https://ai.meta.com/research/impact/fastmri/)
Collected and cleaned large volumes of brain MRI raw data for the fastMRI brain challenge. Created data processing pipelines that perform automatic removal of sensitive patient data (identifiable images).
Reconstructed images from the collected raw data. Provided the reconstructed images to the manual labeling team. Assisted the labeling teem to verify sensitive data removal and labeling.
2018 — 2018
Beijing City, China
Breast Lesion Detection with Deep Learning. (Python)
Lesion Detection software Development: Built and trained models (faster R-CNN based object detection models, MXNet implementation) to output lesion locations and types from input mammogram images.
Worked with data labeling team interactively to make sure the training data are optimized for the models to learn. Actively adopting newly labeled data into training routine and provided feedback to the labeling team based on model learning progress.
Education
2017 — 2020
New York University
Master of Philosophy - MPhil
2017 — 2020
2015 — 2017
University of Michigan
Master of Science (M.S.)
2015 — 2017
2010 — 2014
Huazhong University of Science and Technology
Bachelor of Engineering in Bioinformation Technology
2010 — 2014