Experienced in developing scalable data and AI solutions with a strong foundation in software engineering.
Skilled in building data pipelines, deploying machine learning models, and leveraging cloud infrastructure and
automation to enhance system performance and efficiency.
June 2024 - Present
Machine Learning Engineer
Netflix
I engineered scalable machine learning infrastructure with a focus on real-time personalization and efficient model deployment. This included building a serverless feature store on AWS (Lambda, DynamoDB, S3) to centralize user embeddings, cutting feature duplication by 35%, and enabling EventBridge-driven updates. I optimized large-scale ML pipelines using PySpark and Ray, reducing training runtime by 45% on 1TB+ datasets.
My work in natural language processing involved BERT and spaCy for multilingual content, improving title relevance by 15%. I also implemented LLM-powered content enrichment using GPT-4 and Claude, driving an 18% increase in CTR for long-tail content.
To enhance experimentation and scalability, I built a multi-armed bandit framework for A/B testing and deployed containerized models with Docker, Kubernetes, and CI/CD via Jenkins. Additionally, I developed real-time feature pipelines with Apache Flink and Kafka, delivering sub-500ms latency personalization for 10M+ users.
June 2021 -July 2023
Data Scientist
IBM
I designed and implemented ETL pipelines using Python, SQL, and Apache Spark, improving data processing efficiency by 25% and accelerating model training across enterprise-scale ML workflows. I built and deployed scalable machine learning models on IBM Cloud using Watson Studio, Watson Machine Learning, and Cloud Functions, enabling real-time inference for high-demand applications. By automating the ML lifecycle with Apache Airflow, I reduced manual intervention by 60% through custom DAGs for ingestion, feature engineering, and retraining.
Additionally, I streamlined CI/CD for ML deployment with Git, Docker, Kubernetes, and Jenkins, cutting deployment time by 40% and boosting experimentation velocity. My work also included developing supervised (XGBoost, LightGBM) and unsupervised models (K-Means, DBSCAN) for use cases like fraud detection, customer segmentation, and predictive maintenance—achieving a 30% improvement in anomaly detection accuracy and a 22% drop in false positives. I led the end-to-end SDLC for ML systems, integrating big data tools like Hadoop, Apache Spark, and IBM Cloud Pak to deliver robust solutions in production.
March 2024 -May 2025
Course Facilitator
University of Colorado, Boulder
Facilitated three advanced courses in Statistical Methods and Applications, mentoring over 200 students in key
areas such as exploratory data analysis, probability theory, statistical modeling, and ethical considerations
in data science. Emphasized the development of reproducible statistical workflows and the application of data
science techniques to real-world domains like business and climate science. Instruction focused on hypothesis
testing, p-value interpretation, and the critical evaluation of statistical methods. Additionally, promoted
effective collaboration through self-reflection, peer feedback, and video analysis, while training students
to communicate technical results to non-technical audiences and uphold ethical standards in professional practice.
June 2020 - March 2021
Machine Learning Research Assistant
Aurora’s Degree College
As a Machine Learning Research Assistant at Aurora’s Degree College, I developed a deep learning model using a two-stage
convolutional neural network (CNN) in PyTorch to classify melanoma tissue images, achieving 85% accuracy across six mutation types.
I built an image processing pipeline to handle over 25,000 images and trained models on Google Cloud, improving AUC by 5% through
test-time augmentation. I analyzed misclassifications using color-coded tessellations to better understand prediction errors.
Based on these findings, I extended the model’s capability to classify human melanoma tissue, enhancing its relevance for real-world medical
imaging applications.
