OER Integration at UC San Diego
Institutional research and infrastructure planning for Open Educational Resource adoption, translating equity goals into actionable systems recommendations.
Machine Learning + Space
Production ML Systems · Data Pipelines · Cloud Infrastructure
End-to-end ML systems: data, training, deployment, maintenance
I design and deploy production ML systems with experience in physics-informed modeling, large-scale data integration, and cloud infrastructure. Background in astrophysics and space-adjacent engineering.
Focus Areas
- Production ML systems & automation pipelines
- Scientific ML with simulation-based training
- Large-scale data integration & statistical analysis
- Infrastructure systems & institutional research
Core ML Stack
Python · PyTorch · NumPy · SciPy
AWS (EC2, S3, Batch) · Docker · Linux · HPC
Data Pipelines · Model Evaluation · Automation
Experience
UCSD Astrophysics · ML for GW analysis
Simulation-based training and uncertainty modeling
General Atomics · AI infrastructure
Hybrid on-prem + cloud ML systems
SPACES UCSD · Data-driven ops at scale
Analytics and automation for 890+ student program
Featured Projects
Astronomy Club Newsletter Automation
Production ML pipeline | Weekly automation | Python, APIs, LLMs | Zero-touch ops
Designed a zero-touch, weekly production pipeline using Python + APIs + LLMs, running unattended with monitoring and failure recovery. Reduced manual operations time to zero while maintaining editorial oversight through automated draft generation. Orchestrates 5+ external services (Calendar, Gmail, RSS, Weather APIs) with graceful degradation.
Read More →JWST Gas Morphology Classifier
Physics-informed CNN | Simulated JWST data | PyTorch | 92% accuracy
Built end-to-end PyTorch pipeline for classifying 5 distinct gas morphologies in simulated JWST 4-channel NIRCam data, achieving 92% accuracy (Macro-F1=0.91). Physics-informed augmentation (PSF convolution, viewing angles, detector noise) improved performance by 6.7%. Designed for transfer learning to real JWST observations with interpretability tools for failure analysis.
Read More →Renewable Energy ROI Analysis
Multi-source data pipeline | Federal datasets | Python, statistical modeling | 50-state analysis
Integrated 50-state economic data from multiple federal sources (EIA, Census, NREL) into unified analysis pipeline. Built custom feasibility scoring models combining resource potential, infrastructure costs, and economic factors. Delivered actionable state-level rankings with reproducible methodology and statistical validation.
Read More →OER Integration at UC San Diego
Systems research | Infrastructure design | Cross-functional coordination | Enterprise scale
Led institutional research translating equity goals into technical infrastructure requirements for campus-wide OER adoption. Coordinated across 4 departments (IT, Library, Academic Affairs, Student Services) to design scalable integration with existing LMS and procurement systems. Delivered actionable roadmap with implementation timeline and resource requirements.
Read More →Recent Posts
JWST Gas Morphology Classifier for Binary Evolution
PyTorch-based CNN pipeline for classifying gas structures around binary stars using simulated JWST observations, achieving 92% accuracy with physics-informed...
Astronomy Club Newsletter Automation
Fully automated communications pipeline with LLM-based news curation, weather forecasting, and event aggregation, designed for production reliability and ope...
Renewable Energy ROI Analysis Across U.S. States
Large-scale data integration pipeline evaluating economic ROI of state-level renewable energy transitions, combining federal datasets with custom feasibility...