James Liang Email & Phone Number

Montreal, Quebec, Canada

Working on Flow Production Tracking (formerly ShotGrid).

Montreal, Quebec, Canada

Developing cloud-native technology in Go to automatically manage the deployment, installation, and update lifecycle of cloud-native functions built from Kubernetes and Helmfiles.

Montreal, Quebec, Canada

Developed OCRx, a platform to standardize drug terminologies (drug ontology)• First author of ”Technical support services for drug-related information and interoperability” at MedInfo 2023• Built a full-stack web application using Flask, React, Spring Boot, GraphDB, and Docker to retrieve more than25k+ standardized clinical concepts under 500ms such as medication component, form, and route of administration• Aligned 4+ national drug databases to OCRx’s ontology totalling 76k+ entities using RegEx and NLP heuristics• Fine-tuned RoBERTa on 100k+ drug labels to identify entities within clinical labels with a 95% accuracy

Hosted MariHacks 2023, Montreal’s largest CEGEP hackathon, for the first time in-person at Marianopolis College, raised 4.5k+$ through 7 sponsors and welcomed 178 participants for 2 days.

Montreal

Significantly reduce image noise and data usage in video calls using deep learning, image to image translation of spectrograms, and speech recognition• Reviewed literature on audio resolution and used Wav2Vec to perform real-time speech recognition.• Built personalized speech synthesizer using Tacotron and speaker embeddings generated from 15 second audio clip.• Preprocessed data using Pytorch and audiomentations to acquire Mel Spectrograms and noisy versions of audio.• Worked on using MaskCycleGan-VC to generate robust speech from noisy audio and wav2vec embeddings.• Built a website using React and Flask where users can upload audio and download the denoised version.

Montreal, Quebec, Canada

• Optimized marketing email schedule using historical data to increase click through rates and open rates.• Preprocessed dataset using R scripts and Pandas from raw, untabulated data.• Compared performance of various machine learning models such as XGBoost, Random Forest.• Used state of the art machine learning methods (multitarget encoding, Linear Discriminant Analysis, SMOTE) to predict optimal send time.• Predicted optimal hour range with a top 3 accuracy of 80% (2.6 times higher than baseline model).

Montreal, Quebec, Canada

• Programmed an RShiny app that allows users to create complex SQL-like queries through a simple interface to query 211k+ clinical trials under 2 seconds from clinicaltrials.gov• Leveraged NLP and Google’s Tesseract OCR to extract consent indicators from 17+ years of informed consents• Contributed to Synthea, an open-source synthetic patient generator, by generating localized (Canadian) demographics data for 1854 hospitals and 967 counties (using Statistics Canada open datasets)

Bachelor Of Science - Bs, Honours Mathematics And Computer Science

Activities and Societies: Honours Analysis 2, Honours Algebra 2, Honours Data Structures and Algorithms, Honours ODE GPA: 4.0

Diploma Of College Studies, Honours Pure And Applied Sciences

Activities and Societies: Founding President of the Marianopolis AI Club, where we host weekly workshops on Python, machine learning.

High School Diploma