Experience

Machine Learning Intern · Boston, MA

• —Lead a multi-institutional collaboration between BIDMC and Johns Hopkins to develop and integrate a learning-based autonomous wound closure algorithm within the da Vinci Research Kit (dVRK) platform for future clinical translation, coordinating requirements and experiment milestones across teams
• —Build repeatable wound-closure experiments by standardizing task setup, ROS data capture, labeling guidelines, and evaluation metrics; maintain consistent dataset formats across runs to enable clean comparisons and faster iteration on autonomy behavior
• —Maintain a reproducible NVIDIA Isaac Sim + ORBIT-Surgical pipeline on BU's HPC Cluster using Singularity containers, and document protocols, results, and meeting action items to keep development aligned

Machine Learning Research Assistant · Boston, MA

• —Build end-to-end Python pipelines for POES and OMNI time-series data, generating model-ready datasets with 7M+ 2-second satellite samples per month by ingesting NetCDF files, interpolating gaps, enforcing L-shell (3–9) and MLT quality cuts, and joining with 4-hour windows of 1-minute solar-wind drivers
• —Improve E4_0/E4_90 energetic electron precipitation forecasts, lowering test MAE/RMSE and boosting correlation vs prior baselines, by training Keras neural nets and Ridge regressors with standardized features and time-ordered train/val/test splits
• —Increase interpretability of space-weather predictions across different storm conditions and locations in the radiation belts, improving generalization during rare, high-flux events, by testing different time resolutions (2-second vs 1-minute data), quantile-based flux aggregations, and simple features that capture local time around Earth

IT Consultant

• —Engineered and deployed secure authentication systems using Azure Active Directory and Windows Hello for Business, increasing login efficiency by 25%
• —Implemented Cloud Kerberos Trust with domain controllers, streamlining credential workflows by 40%
• —Built Python scripts to automate data collection, cutting manual tracking by 50% and improving real-time system monitoring

Instructor

• —Delivered engaging math instruction to classes of up to 10 students using diverse methods, raising comprehension by 20% and improving overall engagement by 25%
• —Evaluated 100+ assignments monthly, providing constructive feedback, clear guidelines, and tailored reinforcement to build confidence, discipline, and student growth