Profile
Resourceful machine learning engineer with a post production background. Experienced in statistical analysis, deep learning, NLP, and time series modeling with a strong interest in innovative mission driven machine learning projects and FOSS development.
Skills
General
- Astro
- Cloud Services
- Docker
- HTML and CSS
- Linux
- MySQL
- OOP
- PostgreSQL
- Regex
Python
- Beautiful Soup
- Flask
- GDAL
- Geopandas
- Matplotlib
- NLTK
- Numpyv
- Pandas
- PyTorch
- Requests
- SQLalchemy
- SciPy
- Seaborn
- Selenium
- Shapely
- Statsmodels
- Word2Vec
- scikit-learn
Creative
- Blackmagic Fusion
- Blender
- ComfyUI
- ControlNet
- DaVinci Resolve
- Inkscape
- Krita
- Stable Diffusion
Languages
- English
- Japanese
- Spanish
Experience
Machine Learning Engineer at Virtue Foundation (remote)
From Jan 2022 to Jul 2023 at Virtue Foundation, New York, NY- Created geospatial data pipeline scaled to integrate publicly available GIS data from 72 countries including tiled raster satellite data, vectors from open street maps, internal hospital data from the foundation, etc. to an open source hospital statistical analysis library (Accessmod).
- Deployed the pipeline using a cloud service, Flask API and Docker images for separate Python and R codebases.
- Automated sending the outputs and cached intermediate files to Microsoft Azure based data lake storage.
- Integrated custom logging into the data lake that tracked errors, peak ram usage and per run server costs.
- Found creative solutions to enrich existing datasets including a random forest regressor to impute gaps in health facility data and a workflow to sample from the Google directions API to inform better estimates of road speeds.
- Automated a geospatial workflow to standardize all inputs to the same equal area projection and resolution with minimal loss and distortion generating custom projection strings based on a region’s centroid coordinates.
- Built custom sanity checks to measure result correlation to development and poverty indicators such as data from global health surveys and nighttime satellite illumination data.
- Created custom interactive visualizations using the raw output data. Formulated a flexible and efficient way to vectorize raster outputs in a clear way in order to work with the Carto platform.
Data Science Intern at Virtue Foundation (remote)
From Sep 2020 to May 2021 at Virtue Foundation, New York, NY- Optimized scraping pipelines for gathering millions of tweets and their metadata in English speaking developing countries.
- Established procedures to clean and encode unstructured data (including emojis) for supervised learning using chains of regular expressions along with TFIDF and word2vec representations to identify relevance to healthcare topics.
- Created custom up to date word embeddings that utilizing custom regex to detect emoticon combinations in addition to embeddings for various emojis using scraped data and libraries from Stanford’s GloVe research.
- Drafted workflows for several unsupervised techniques including K-means clustering and Latent Dirichlet Allocation.
- Implemented a custom aggregation strategy benchmarked in an academic publication to improve vector representation of a short body of text.
- Enhanced integrity of the data by integrating filters based on additional metrics from user data.
- Improved the data pipeline by migrating existing data to a PostgreSQL Database for improved efficiency.
Freelance Post Production Artist
From Jul 2015 to Mar 2021 at Multiple Clients, New York, NY- Assembled visual content for Boy Wonder Productions on over 10 seasons of TV shows on HGTV and the DIY channel.
- Handled tasks within the post-production pipeline, including motion graphics, compositing, and color correction.
- Crafted content for a diverse set of clients and projects (assuming a multitude of post production roles) across independent films, fashion, and short form interviews.
Education