Data Origin: Researchers in the Atomic-Molecular-&-Optical-Sciences Group, within the Chemical Sciences Division at Berkeley Lab (LBNL), have developed a novel technique for the investigation of nonlinear physical processes with time-resolved measurements scaled at the molecular level.

I am passionate about building data visualizations that inspire and inform. Most recently I built an analytics dashboard for a research team at Lawrence Berkeley National Laboratory, with Plotly and Dash for Python. As a visually appealing web application, it was recently featured by Plotly on Twitter after receiving positive feedback from the Community Forum. To see my freelance contributions bring these two powerful technical communities together was gratifying. I had often fantasized about building interactive visualizations of my own experimental research, as a graduate student in physics, but back then I lacked the Python proficiency to do so. Overall this was an important project because I was able to successfully mimic timeless features of legacy data visualization software, but with the added convenience and flexibility inherent to open-source technology.

The application is composed of five Python files structured in a modular format. It can be launched via Heroku, but as an analytics tool it was designed to be driven primarily through a Jupyter Notebook. It scans a shared directory for raw data stored in spreadsheets(.csv, .tsv, .dat, etc.) and imports the Pandas library to conduct preprocessing and build out a reporting pipeline.

Fun Fact: This is a screenshot of heatmaps that display hardware generated data as a single experiment, or as an average of many experiments. The voltage signal originates from a set of three photodiodes and is mapped into a continuous color palette for visual analysis. A Vertical Range Slider, along the right hand side of each figure(in light blue), provides dynamic control over the range of frequencies scaled.

For each dataset in the folder an automated report is generated providing users with a preliminary view of the data quality, as well as with a descriptive summary of its content. The team was seeking a way to integrate data from multiple sources and, through a dropdown menu, each dataset in the specified directory can be selected for either individual analysis or a side-by-side comparison.

Fun Fact: In this screenshot a number of Display Options can be seen, such as dynamic control over legend colors, that are meant to support researchers performing analysis in real-time. The laboratory safety team, at the facility, requires that researchers wear protective eyewear while the laser is in operation. This effectively means specific frequencies of light are harder to see than others while the team is collecting their data.

This was a satisfying product to deliver because it streamlined the research team’s ability to collaborate and perform analysis. They can now explore the data together, from multiple workstations, at the same time. Additionally, my decision to build out the components with the Dash-Bootstrap library resulted in a responsive layout, allowing for impromptu analysis to be conducted on handheld mobile devices. Furthermore, the use of Python allows the app to evolve along with the needs of the team, so they can focus on their unique mission and not get held back by the cost of proprietary software.

For more information check out the project hosted on GitHub. The README details more technical aspects of the dashboard and also provides reference to an overview of the science. For examples of additional data projects I’ve enjoyed working on, check out my portfolio! Thanks for reading and please feel free to ask questions, or provide feedback, in the comments section below.