Machine Learning Engineer  • Binflux Inc.

Oct 2021 - Now

• Improved 80% efficiency and reduced manual intervention by designing and building ETL workflows and batch processing pipelines from scratch for AI products using Airflow and Python.

• Reduced data processing time by 50% through implementing incremental data loading to optimize ETL processes.

• Reduced data errors by 75% through implementing data validation to optimize data pipelines.

• Designed and built a deep learning pipeline from scratch to monitor embryos using Python, MLflow, Hydra, Pytorch Lightning.

  • Reached 99.9% accuracy, comparable to SOTA.

  • Improved 10% accuracy in the more challenging scenario by further introducing new learning methodologies.

• Improved a deep/machine learning pipeline to conduct embryo selection using Python, MLflow, Hydra, Tensorflow.

  • Increased 10% AUC and generalizability by combining heterogeneous data. 

  • Improved 2x data privacy and security by introducing decentralized learning.

  • Reduced 66% manual intervention by building a one-command decentralized training pipeline using Ansible.

• Built and maintained AI services of previous pipelines using FastAPI, Tensorflow Serving, MariaDB, MongoDB, Redis, Docker and Docker Compose.

• Edited and wrote 3 papers for journal publication: Using an Interpretable Machine Learning Model to Predict Corifollitropin Alfa Protocol, and 2 WIP.