Waterloo, Ontario, Canada
• Developed a log anomaly detection platform combining ML, data pipelines, CI/CD, APIs & a NoSQL database
• Researched NLP anomaly detection techniques & productionized an unsupervised ensemble model leveraging LSTM/Dense autoencoders, Google’s BERT transformer, and scikit-learn models (IsolationForest, PCA, etc.)
• Used Tensorflow, Pandas, and Numpy for model development and achieved high F1-score & accuracy
• Improved model inferencing by 38,250% & improved data pipelines by removing redundant logging and API calls
• Optimized low level memory usage by 52% via memory tracing, profiling, implementing subprocesses (multiprocessing)
• Implemented CI/CD using GitLab Runner and Docker to automate tests ensuring successful integration and deployment
