Ali Murad

• Led a team of 5 Data Scientists to develop Data Science Computer Vision use cases in Radiology and Natural Language Processing use cases with Generative AI and Large Language Models (LLMs).
• Built infrastructure in Google Cloud to combine HCA's repository of 30 billion radiology images with radiologist impressions to automate data labeling for supervised learning.
• Led the development of Nurse Shift Handoff report generated using Large Language Models. Applied zero-shot and few-shot learning and model fine tuning to optimize model outputs. This report is utilized by nurses at the time of their shift change to transfer care of patient to the incoming nurse.
• Led the development of Emergency Department Transfer Notes generated using Large Language Models. These notes are utilized by physicians to provide a plan of care as they discharge patients from the emergency department.
• Led the development of radiologist work prioritization models for Head Stroke and Pulmonary Embolism Computed Tomography (CT Scans). Due to the national shortage of radiologists reading radiology studies have a long delays which can be detrimental for patients in life threatening cases. These prioritization tools help radiologists determine which radiology studies have a high level of severity so they can be read in a timely manner.

• Led the data science Google Cloud computing strategy as part of the $650 million investment by HCA Healthcare in Google Cloud.
• Worked with the cloud infrastructure team and utilized Terraform Infrastructure as Code to create and test cloud infrastructure for the data science teams across HCA.
• Utilized Google Cloud Platform with Vertex AI, Google Cloud Storage, Artifact Registry, Cloud Run, Big Query, and Healthcare API to build and test applications for patient care.
• Built autoregressive forecasting models and improved patient forecasting by 35% to optimize nurse staffing.
• Utilized GitHub to version control code and built CI/CD pipelines for model deployment.

• Worked on building statistical time series forecast models for patient volume prediction.
• Built and optimized deep learning computer vision models on Chest Xrays for diagnosis of thoracic diseases in patients.
• Used Python with scientific computing libraries including Sklearn, TensorFlow, and Statsmodels to build and test machine learning and statistical models.
• Used Openshift, Docker, and Google Cloud for model deployment and integration.
• Used JIRA and GitHub to organize workflow and maintain project code base.

• Built and maintained statistical time series models, in Python, to predict patient volume and staffing requirements on a monthly and yearly basis.
• Built dashboards to visualize patient volume at all serviced hospitals using SQL and Power BI.
• Built ad hoc reports, in SQL, Teradata, Python, and SAP Business Objects, to be provided to different departments in order to facilitate their workflow.
• Used SQL and Python to implement a data extraction and loading procedure for the Prebill Denials department and reduced latency by 6 hours.
• Implemented a forecasting solution using time series models, in Python, to predict the volume of calls received at the customer service center to aid in estimating staff requirements.
• Built and analyzed monthly operating reviews to be used by the executive team, for all serviced hospitals.
• Mentored a financial analyst in training by collaborating on ad hoc requests and projects.

- Classifier for Breast Cancer Classification with scikit-learn. Models used Logistic Regression.

o Tools used: Pandas, Numpy, Matplotlib, and Scikit-learn.

- Did data pre-processing and built a seasonal autoregressive model to forecast the amount of SO2 in air.

o Tools used: Pandas, Numpy, Scikit-learn, Statsmodels, Pyramid Arima

- Built a Natural Language Processing application to rate hotel reviews by users in real time.

o Tools used: Python, NLTK, Pandas, Numpy, Scikit-learn, Heroku, Flask, and Matplotlib

- Built a movie recommender system by using natural language processing. Designed a project pipeline including data storage, data pre-processing, feature engineering and selection, and modeling.

o Tools used: Pandas, Numpy, SQLite, Scikit-learn, Natural Language Toolkit NLTK, Matplotlib.

- Built a tweet classifier to classify if a given tweet based on its text, location, and keywords signifies a disaster/emergency or not. Designed a project pipeline including data storage and retrieval, data pre-processing, feature engineering and selection, and modeling.

o Tools used: Pandas, Numpy, SQLite, Scikit-learn, Natural Language Toolkit NLTK, Matplotlib.

- Did data pre-processing, feature engineering, and model building and evaluation on the Lending Club loan data to classify defaulted loans. Models used include Logistic Regression, Naïve Bayes, Random Forest, and K-Nearest Neighbors

o Tools used: Pandas, Numpy, Scikit-learn, Matplotlib, and Seaborn.