Machine Learning Engineer at Radar Co

Developed a neural network for human-presence detection from radar point-cloud data, achieving 95% accuracy. Built an automated annotation pipeline deployed on AWS (Lambda, EC2, S3) to generate labeled radar datasets at scale. Implemented spatiotemporal clustering and target association using scikit-learn, leveraging spatial and doppler features to reduce false positives and improve track stability.

Research Assistant at MIT ARCLab

Developed ConvGRU architecture from scratch in Python/PyTorch for long-term space object density forecasting, achieving 100x speedup over traditional physics-based methods and accurate 100-year forecasting capability. Fine-tuned the model for improved generalization across diverse orbital scenarios, integrating a guidance mechanism that reduced long-term forecast error by 30%. This resulted in a peer-reviewed paper.

Undergraduate Research Assistant at MIT Senseable City Lab

Developed a custom data extraction pipeline in Python, integrating Google Maps APIs and weather data to capture high-resolution image and temperature datasets. Implemented an end-to-end inference pipeline with semantic segmentation models in Python/PyTorch, enabling automated urban environmental analysis.