* Startup, Stealth Mode (Financial Services) [from 4/2013]: Directed, coordinated and individually implemented full stack product development from initial bootstrapped product sketch to seed stage funding prototype.
Built various API adapters, sanitizers and scrapers for serializing external data into normalized relational datastore using Django.
Architected, implemented, operated automated testing infrastructure and deployment for a RESTful Write-heavy Python API backend, which used OAuth signup/authentication, social account provisioning and social profile importation to provide time range based aggregation and filtering through data modeling and API provisioning with Django, PostgreSQL and bit of caching.
Architected and deployed asynchronous computation platform to periodically discover, update and integrate user information from 3rd-party social API sources for users both initially and as time continued to go on.
Designed, prototyped, and interfaced minimalistic UI/UX for time-series frontend with backend API.
Instrumented basic Cloud provisioning and deployment.
* VC backed startup (Computer Software) [from 6/2012]: Directed migration of Django code from MySQL to PostgreSQL, as well as migration from raw SQL to ORM using code.
Directed migration of version control management from private repository to GitHub repository as well as adoption of GitHub issues, branch creation and pull request practices.
Developed components which import, serialize and use external Facebook and Twitter API data for analysis -- used Facebook Graph API and FQL api, along with streaming Twitter API.
Participated in lean product development from initial sketches of ideas to backend architecture.
* Funded startup (Computer Software) [from 12/2011]: Extended and refined Django backend.
Rehauled and extended unit tests of API importer app internals in Django, as well as unit tests for API behavior and permissions (using the TastyPie API for Django).
Built various API adapters, sanitizers and scrapers for serializing external data into normalized relational datastore (PostgreSQL) using the Django ORM.
* Bard College at Simon's Rock (Computer Software) [from 5/2011]: Conducted research for senior thesis on data mining software repositories and other project history of open source software projects, especially those foundational to LAMP-like *NIX web application stacks.
Successfully deployed platform on Amazon's EC2 cloud to construct decision trees predicting most frequent directory activity during some period of revision history from email address of most frequent contributor and plot a graphical representation of these trees using matplotlib and NetworkX in Python.
* University of Chicago Medical Center (Hospital & Health Care) [from 8/2010]: Conducted research on computer aided diagnosis of breast cancer as a summer fellow through AAPM's MUSE summer fellowship.
Used MATLAB to explore behavior of machine learning classifiers under different forms of dimension reduction on simulated clinical lesion data.
Wrote a tool to visualize performance of Bayesian classifier. The tool color mapped test cases as points in 2D space based on inferred class, and allowed for finer interrogation of any point case of interest. The tool plotted the full inferred probability distribution from the classifier for any case of interest, allowing for a visual representation of the the neural network's confidence of distinguishing the class of that point.
Explored configuration space of classifiers and dimension reducers for those yielding optimal classification performance.
Compared performance of multiple forms of dimension reduction across a diverse corpus of data to explore reduction and classification behavior by helping run machine learning trials across a broad set of variably shaped data, particularly varying in feature space dimension and size of training set.
Used PostgreSQL, matplotlib and Python to store, analyze and graph interesting results from the simulations.
Results from these simulations crystallized into SPIE Medical Imaging 2011 conference proceedings.
* Wesleyan University (Higher Education) [from 8/2007]: Used LABVIEW to implement stepper motor instrumentation for controlling laser orientation during experiments.
* Yale University School of Medicine (Research) [from 8/2006]: Coded image processing tools in MATLAB used for processing experimental MRI data at Yale's MRRC by combining data from multiple axes.
Began construction of voxel based signal interrogation tool in MATLAB which plotted signal intensities of a matrix of signal alongside the signal trace of any signal of interest in particular.
Conducted noise simulations using MATLAB to investigate the effect of high signal to noise ratios to low ones, related to the effect of noise on in vivo versus in vitro NMR spectra.