Synchrony Financial // AI Systems Intern, Credit Modeling
Sept 2026 – Present
- Strengthened credit model governance across a multi-million-account lending portfolio by building automated PySpark data validation and drift-detection pipelines, flagging 15% of low-confidence model regions ahead of quarterly governance review and cutting manual QA effort by 30%.
- Accelerated loss forecasting and model health reporting for cross-functional stakeholders by engineering Python/PySpark ETL pipelines across delinquency, payment, and sales data and building recurring Tableau dashboards, reducing reporting turnaround by 40%.
PySparkPythondrift detectionETLTableau
John Deere // Machine Learning Intern, Self-Driving Combines
Sept 2025 – Aug 2026
- Deployed an ML artifact removal pipeline to 10K+ farmers using PyTorch, reducing yield map noise by 73%, validated through a 20-customer panel.
- Designed and implemented a self-supervised learning quality metric adopted company-wide as a customer-facing standard, enabling automated scoring and flagging of low-confidence regions.
- Developed computer vision and reinforcement learning features for autonomous combine harvesters, integrating multi-sensor fusion for real-time obstacle detection and harvesting path optimization.
- Translated complex analytical findings into reports and presentations, communicating technical insights to key stakeholders to inform business decisions.
PyTorchtransformersreinforcement learningcomputer visionsensor fusion
MIT Lincoln Laboratory // Software Engineering (AI) Intern
May 2024 – June 2025
- Engineered a deep learning pipeline to classify 3D printing failures using a curated 10K-image subset of the CAXTON dataset.
- Benchmarked ResNet, ViT, and Swin models with ResNet50 achieving 89% accuracy; used feature extraction to cut model size by 99.9%.
- Resolved data quality issues and conducted model analysis to inform autonomous defect detection in additive manufacturing.
- Fine-tuned BERT and integrated SVMs using PyTorch and Hugging Face to build a robust fake news classifier.
- Benchmarked CNNs, RNNs, and transformers to optimize performance and generalize misinformation detection across sources.
computer visionNLPBERTanomaly detectionHugging Face