
AI Software Engineer
Saigeware Technologies · Feb 2023 – Present
- Developed a cohort of 120,000 cardiovascular and cerebrovascular patients from Mayo Clinic's EMR database, resulting in 310,000 admissions.
- Utilized SQL to implement inclusion and exclusion criteria aligned with business requirements.
- Developed an XGBoost-based readmission risk model with ~82% accuracy and precision (0.74) and recall (0.75).
- Significantly outperformed LACE plus scores in predicting both short-term (30-day) and long-term (6-month) readmissions.
- Developed an alert system to monitor patient vitals and detect out-of-range values using a window-based outlier detection algorithm.
- Implemented automated notifications to alert patients and providers of abnormal readings.
- Developed a baseline risk prediction model using phenotypic markers from GT+.
- Trained the model with TensorFlow and XGBoost to predict patient susceptibility to specific risks based on collected marker
- Developed physical and mental wellness scores by integrating health data (steps, calories, SPO2, heart rate, HRV, and voice).
- Utilized fuzzy logic algorithms to model and quantify wellness levels.
- Designed a risk clustering model to determine the main causes of hearing loss and tinnitus in patients.
- Implemented a separate model to predict mental wellness outcomes, including depression, burnout, and anxiety, and their progression over time.
- Implemented a data quality assurance system to evaluate input data and detect drift over time.
- The system also facilitated monitoring of machine learning model performance in production.
- Used YOLOv7 and YOLOv8 to extract and segment visual data, feeding the results into the baseline risk prediction pipeline
- Automated the generation of PDF reports using Python for the Remote Patient Monitoring (RPM) program.
- Reports include patient vitals, episode-based summaries via ChatGPT API, and outlier analysis, which are e-faxed to providers for insurance claims.
- Creating agents using Google ADK and Langchain to generate Audiological insights using various Audiological test results and demographical data.
- Creating a vector store using Chroma DB to retrieve data that relates to Audiological conditions and patient outcomes such as mental wellness, cognitive wellness and auditory health.

