These past months I've spent most of my time working, with my responsibilities ranging from consolidating our dockerized development frontend onto one Nginx server, to a ridiculous amount of two-line CSS changes. Full stack development with a CI/CD schedule means you never get bored. A week might include: * resolving Redis cache race conditions due to Django's deletion model, * fixing a complex checkout edge case, * discovering a timezone issue after tracking it through 4 different logging systems, * hammering out the specs to a new nighttime routing React app with Product. A highlight reel: * implemented driver tips, web notifications, credit card deletion, and a new mobile item view, * refactored checkout flow, integrated Jupyter notebooks into Django, worked on custom menus (BYO Meals), * debugged production issues while on-call I crazy enjoyed this job, and the challenges arising from this interconnected web of API calls and database updates and half-implemented interfaces. There's always an intractable problem still left to think about, and that's what makes it so fun. I left the company during a company-wide layoff.

At AuriQ, I helped a small team deliver a scalable software platform (Essentia) for big data management and analysis (ETL). Connecting to data in cloud storage solutions, Essentia made categorizing, querying, preprocessing, and analysis very easy, with unlimited cloud scaling built-in. * developed software to quickly scan, categorize, and preprocess a variety of streamed log files from S3 buckets using Python and Bash scripting, * designed linked MySQL databases to hold generated metadata with SQLite summary files, * ran machine learning user classifications with decision trees and multinominal logistic regression to generate client recommendations, * worked on several visualizations and presentations of client conversions, both using Ipython/Matplotlib/Pandas and managing data pipeline to D3.js graphs, * requisitioned cloud computing resources (AWS) This was my first foray into industry, and it was an amazingly instructive and fun experience, with interesting datascience problems, cloud computing software, and databases. I worked full time during the first half year, then part-time for a few months while I finished 2 classes at Caltech. I left the company when I graduated. Essentia

Dorm/House Waiter for two school years.

Power spectra from molecular cloud images show an apparent characteristic width for filaments where most, if not all, protostellar cores are found. However, previous observations of different clouds did not have this signature. By simulating model clouds, I attempted to lose the power spectra signal that implies a characteristic filament width, through randomization and adjustment of several input parameters. * Python simulation of molecular cloud filaments to theoretically test a power spectra discrepancy reported from the Herschel telescope * interpolated spectra, ran FFTs, and refined heuristics to pare down the variable exploration space * discovered a spectra anomaly that could lead to a false positive for a characteristic filament width * observing run at the Skinikas Observatory, Crete * work with the international science community Filaments in Molecular Clouds

I managed a hardware-integrated data pipeline project for a new telescope array. MINERVA is an array of 6 small-aperture robotic telescopes used for exoplanetary science capable of both radial velocimetry and transit studies. Ground-based telescopes require protection from the ground-level thermal contribution to astronomical seeing. I was tasked with a complete understanding of the microtubulence and wind vectors, temperature profile, vibrational stability, and seeing at four possible sites. My report (below) was essential to my mentor’s confident deployment at Whipple Observatory, AZ. * high-frequency timeseries Matlab data pipeline design and management * involved hands-on work in shop and construction of enclosures to protect dataloggers, thermocouples, and more * observing time on the Palomar 200’’ and the 60’’ * handling practical experimentation: delicate instrumentation, bureaucratic deadlines, variable weather conditions Technical notes on the MINERVA Project Sites

I optimized the California Planet Search (CPS) for multibody exoplanetary systems on the Keck telescopes, using Python Markov Chain Monte Carlo (MCMC) simulations and Bayesian analysis for prioritization of the observing strategy. Also theoretically limited outer companion signatures with a minimum mass scaling relationship dependent on inner planet radial velocity dispersions, for further study of migration mechanisms.

Weekly lessons to neighborhood kids