Yubo Sun, Phd Email & Phone Number

California, United States

Burlingame, California, United States

● Extracted time series data on the Caban hybrid energy system load energy usage through a telemetry API, constructing a Long Short-Term Memory (LSTM) model in TensorFlow within the Jupyter Notebook. Achieved an R^2 of approximately 0.98 for forecasting load energy consumption, contributing to smart energy management initiatives such as peak load shifting and shaving.● Led the development and oversight of the ingress service, implementing an Extract, Load, Transform (ETL) data pipeline… Show more ● Extracted time series data on the Caban hybrid energy system load energy usage through a telemetry API, constructing a Long Short-Term Memory (LSTM) model in TensorFlow within the Jupyter Notebook. Achieved an R^2 of approximately 0.98 for forecasting load energy consumption, contributing to smart energy management initiatives such as peak load shifting and shaving.● Led the development and oversight of the ingress service, implementing an Extract, Load, Transform (ETL) data pipeline deployed on AWS ECS containers. Conducted thorough statistical analyses to establish thresholds for alerts, ensuring robust data handling and accuracy.● Effectively managed activities within the data warehouse (TimescaleDB), encompassing tasks such as creating minute, hourly, and daily aggregates, establishing continuous views, setting retention policies, and executing data backfill actions to address data corrections.● Executed PostgreSQL queries to extract Energy Key Performance Indicators (KPIs) from TimescaleDB, leveraging and visualizing the data to construct an environment dashboard. Established a streamlined workflow for the Network Operations Center (NOC) team by creating workbooks for automated reporting and analytics using the BI tool Sigma.● Engaged in collaborative efforts with the Firmware and Hardware Teams, conducting exploratory data analysis (EDA) to gain insights into device behaviors. Provided consultative services to the hardware team, ensuring the establishment of intelligent alert rules to minimize unnecessary alerts from the systems. Show less

San Francisco Bay Area

• Developed 10+ C language-based algorithms & test sequences on memory device tester to collect erase, program, read and threshold voltage distribution data & parsed ~5GB big data generated from NAND device experiments.• Extracted memory device screening data from Amazon redshift database & visualized it with a sigma plot to perform failure data analysis of memory devices, assessed the strength of stress & screen for memory devices, with 95% confidence.• Implemented data cleaning… Show more • Developed 10+ C language-based algorithms & test sequences on memory device tester to collect erase, program, read and threshold voltage distribution data & parsed ~5GB big data generated from NAND device experiments.• Extracted memory device screening data from Amazon redshift database & visualized it with a sigma plot to perform failure data analysis of memory devices, assessed the strength of stress & screen for memory devices, with 95% confidence.• Implemented data cleaning, filtering, imputation & transformation, evaluated device data quality, consistency & integrity with Python Pandas, NumPy, SciPy & Spotfire Analyzer, identified data anomalies & devised data remediation plans.• Performed data analysis to validate self-proposed memory device parameter changes, applied A/B tests to quantify the impact of the change, presented analytical insights with device team, and communicated with design, system & process team for alternative solutions & strategic decision-making.• Proposed an SVM Machine Learning model for memory holes quality detection, utilized Canny operator to detect edges of memory holes & Hough transform with thresholds to determine circularity of memory holes with 88.3% accuracy & enabled data-driven quality judgements. Show less

Remote

• Scraped ~8GB materials data using manufacturing APIs in Python, performed data mining & wrangling, ETL, imputed missing data, identified & treated outliers using Python & PySpark on AWS.• Applied exploratory data analysis to analyze trends & patterns and feature correlations, visualized them in QuickSight.• Implemented SARIMA time series model, Random Forest & XGBoost regression algorithms in Python using Scikit-learn to predict product prices, incorporated 5-fold cross-validation &… Show more • Scraped ~8GB materials data using manufacturing APIs in Python, performed data mining & wrangling, ETL, imputed missing data, identified & treated outliers using Python & PySpark on AWS.• Applied exploratory data analysis to analyze trends & patterns and feature correlations, visualized them in QuickSight.• Implemented SARIMA time series model, Random Forest & XGBoost regression algorithms in Python using Scikit-learn to predict product prices, incorporated 5-fold cross-validation & built data processing & machine learning pipelines and achieved R-squared value of 0.83 and accuracy of 87%. Show less

West Lafayette, Indiana, United States

Unsupervised Screening and Automated Detection of Photovoltaics Cell Performance • Leveraged Keras in Python to build a Neural Network Autoencoder to encode ~1GB of current-voltage characteristics data generated from TCAD Sentaurus device simulation framework in a lower dimensional space. • Cleaned data, removed anomalous data and applied max-min normalization on delta current-voltage characteristics to transform negative values from the dataset & ensured faster convergence of training… Show more Unsupervised Screening and Automated Detection of Photovoltaics Cell Performance • Leveraged Keras in Python to build a Neural Network Autoencoder to encode ~1GB of current-voltage characteristics data generated from TCAD Sentaurus device simulation framework in a lower dimensional space. • Cleaned data, removed anomalous data and applied max-min normalization on delta current-voltage characteristics to transform negative values from the dataset & ensured faster convergence of training process by 55%. • Performed K-means clustering, applied the elbow method to extract 7 clusters in low dimensional space & identified degraded Photovoltaics Cells along with their correct root cause with accuracy > 95%, results are under review in JPVNumerical & Analytical Model for Rear Heterojunction Gallium Indium Phosphide Solar Cell • Built a TCAD Sentaurus simulation workbench to develop an opto-electric coupled numerical simulation framework to model light absorption & electrical power conversion. • Applied Poisson equation to solve electrical carrier transport phenomenon in an analytical model & implemented polynomial feature engineering to identify key device parameters of highly efficient solar cells. • Published the modeling outcomes in Journal of Photovoltaics. DOI: 10.1109/JPHOTOV.2019.2892530Tailoring Interdigitated Back Contacts for High Performance Silicon Solar Cell• Built 2-D opto-electric coupled simulation in Sentaurus to model Silicon solar cell & benchmarked against experiment data. • Developed a gradient descent algorithm for regression analysis on data generated from the modeling process, predicted an optimum back contact design structure for maximum power output.• Summarized the findings in a research paper & published in Applied Physics Letter, DOI: https://doi.org/10.1063/1.5080075 Show less

Doctor Of Philosophy - Phd, Electrical And Computer Engineering

Bachelor'S Degree Of Science, Electrical And Computer Engineering