Sinke Legesse Lete

CVS Health — AI / ML Engineer

Mar 2023 – Nov 2025

Built and deployed predictive models using XGBoost and CatBoost on large-scale patient datasets in Amazon Redshift, improving forecast accuracy by 18% and generating $1.2M in annual savings.
Designed automated clinical scoring pipelines with AWS SageMaker, enabling scalable model training, validation, and deployment for healthcare applications.
Developed production-grade NLP pipelines using BERT, LangChain, and Hugging Face to summarize patient documents, reducing manual review workload by 24% across 10,000+ weekly records.
Engineered computer vision models (CNN, RNN, LSTM, YOLOv8) on medical imaging data in Amazon S3, improving anomaly detection and diagnostic efficiency by 21%.
Implemented robust data and ML pipelines using Apache Airflow and AWS Glue, processing 200K+ records per day for feature engineering, clustering, and predictive analytics.
Deployed low-latency inference APIs with FastAPI, optimized models via ONNX, and ensured explainability using SHAP and LIME, achieving 99% production uptime.

Northern Trust — AI / ML Software Engineer

May 2021 – Feb 2024

Developed regression and classification models using Python, XGBoost, and CatBoost, improving portfolio risk prediction accuracy by 14% and preventing $2.3M in potential losses.
Built NLP workflows with spaCy, BERT, GPT-4, and NER to automate regulatory document classification, reducing compliance review time by 27% across 1,200+ monthly reports.
Implemented time series forecasting models on historical trading data using Pandas and NumPy, improving predictive accuracy by 17% with Azure ML and Databricks.
Delivered audit-ready model explainability using SHAP, reducing manual reconciliation errors by 22% during portfolio compliance checks.
Automated ML pipelines with Apache Airflow, Docker, Jenkins, and Azure Synapse Analytics, improving reproducibility by 95% and cutting deployment time by 35%.
Designed Power BI dashboards backed by PostgreSQL and MongoDB, enabling real-time insights for 50+ portfolio managers handling 10,000+ daily transactions.

Booz Allen Hamilton — Data Engineer

May 2020 – May 2021

Engineered secure and scalable data workflows in AWS supporting the National Cancer Institute (CTRP) project.
Designed and maintained ETL pipelines using Python, improving data ingestion reliability and downstream analytics quality.
Performed data enrichment, validation, and transformation to support analytics and ML use cases in healthcare and life sciences.
Collaborated with data scientists and ML engineers to productionize ML-ready datasets for modeling and reporting.
Supported ML-driven initiatives for federal clients including the CDC and FDA, ensuring compliance with data governance standards.

Nationwide Insurance — Data Science Intern

Aug 2019 – May 2020

Delivered an end-to-end data science solution for Commercial Lines insurance, generating significant workload savings for the business unit.
Built a scalable Microsoft Bing-based web scraper, performing extensive data cleaning, normalization, and validation.
Implemented NLP semantic matching pipelines using TF-IDF and SIF weighted embeddings for data enrichment.
Developed and tuned a bootstrap bias-corrected, cross-validated XGBoost model to automatically grade insurance risk.
Applied business-driven thresholding strategies to align model outputs with underwriting and policy decision requirements.

About