![Profile](https://weekday-user-pictures.s3.ap-south-1.amazonaws.com/profile-images/haoransh.jpg)
Allen (Haoran) Shi
Sr. Applied Scientist at Amazon, Passionate Machine Learning Practitioner, Einstein Visa Holder
4.7
Years of Experience
Education
carnegie mellon university, peking university
Companies
amazon, carnegie mellon university school of computer science, petuum inc., grab, stanford university, peking university
Reach out to Allen (Haoran) Shi via Email, InMail and SMS drip
by installing Chrome extension
Allen (Haoran)'s contact details
Email (Verified)
haoXXXXXXXXXXXXXXXXom
Experience
2020 - Present
amazon
Senior Applied Scientist
2019 - 2019
carnegie mellon university school of computer science
Graduate Teaching Assistant
TA for Applied Machine Learning by Prof. Carolyn Rose
2019 - 2019
petuum inc.
Machine Learning Intern
• Designed and built the skeleton of NLP-pipeline project which serves as the infrastructure of machine learning solutions, including data packing schema, configuration-based programming, bridging the gap between distributed system team and machine learning teams • Built the CNN-biLSTM-CRF model and Bio-Bert model for name entity recognition (NER) and integrated it with the span-based semantic role labeling (SRL) model in NLP-pipeline framework
2018 - 2018
grab
Backend Engineer Intern
• Designed the pixel-level classification model with CNN for map reconstruction from images containing million-level GPS point traces, automate the process to update the connectivity of road maps which is essential for in-car navigation • Implemented the evaluation algorithm for roadmap quality with topological exploration and visualized the running of missing road detection, providing a more robust metric which captures the geometry and topology of the roadmap quantitatively • Improved the F1 score of missing road detection from 0.4 to 0.7. The pipeline provided an end-to-end solution for online map refinement and unified the working flow of map refinement task for the whole GEO-road team
2017 - 2017
stanford university
Healthcare Innovation Seminar and Hackathon
• Led a multidiscipline team to develop an Integrated app to identify cataracts and refer patients to clinics in remote areas • Designed the business model and proposed the Image Classification model based on ResNet-20 and various regularization techniques • Received Grand Prize at Stanford University Interdisciplinary Hackathon (1/53) and Intel Nervana AI Cluster Runner-Ups Prize (3/53) More information on https://devpost.com/software/cataspot
2016 - 2017
peking university
Undergraduate Research Assistant
Advisor: Prof. Ming Zhang • Representations of Heterogeneous Event Sequences for Clinical Prediction: built heterogeneous event sequence modeling framework for timely clinical prediction by adaptive segmentation and heterogeneous encoding; provided qualitative analysis of the experiments to validate the effectiveness of different modules and proposed improvement strategies • Dynamic Social Recommendation via Neural Attention Model: crawled movie rating and rater social network from douban website and organized a large dataset, then implemented recurrent neural network model for movie recommendation system • Proposed an attention-based deep learning pipeline for International Classification of Diseases (ICD) Coding of text medical documents, with character-aware language encoding, Tree-of-Sequences LSTM for syntactic extraction, attention mechanism, and adversarial learning. Achieved state-of-the-art accuracy and F-1 score on the MIMIC-III dataset. Published a first-author paper on arXiv and a second-author oral paper on ACL 2018.
Experience
9 Skills
Applied Machine Learning
BERT
Convolutional Neural Networks (CNN)
infra
Infrastructure
Machine Learning (ML)
Natural Language Processing (NLP)
Research Scientist
teaching
Education
2018 - 2019
carnegie mellon university
Master of Computational Data Science
Computer Science
2014 - 2018
peking university
Bachelor's
Computer Science