I bring passion and precision to every project. As a patented expert in machine learning and AI, my focus is on crafting user-friendly frontends that showcase the power of ML models. I create solutions that are visually appealing, functionally sound, and perfectly aligned with customer needs.
I’m available for consulting, and just touching base about all things ML & AI, so don’t hesitate to get in touch!
I’m a machine learning engineer and scientist with a passion for product development. I’ve got some management chops too.
I have a background in computer science and physics, with a PhD in Astronomy, and a strong interest in the intersection of the physical sciences and literature.
Here are a few technologies I've been working with recently:I led a team of machine learning engineers and iOS developers to launch incredible AR experiences at Smile Direct Club. We let customers see their future smile using generative AI, along with a bunch of other awesome stuff. Multiple patents obtained.
The shortlist:
GANs to visualize your teeth,
YOLOs to guide the user through a smart-phone based 3d scan,
2d images to 3d meshes at orthodontic accuracy,
ARKit-based training data collection (internal company tool)
The longer version:
Headed the ML engineering team, developing the “Smile Maker Platform”, an ML-powered augmented reality app, significantly impacting corporate growth and sales strategy.
I was product owner for the app. At peak usage the app had >30K MAUs, was ranked as a top-3 Medical iOS app, and doubled the probability for the user to purchase our products.
Led the development and deployment of customer-facing ML models and algorithms across iOS, Android, and Web.
Led a diverse team of 4 scientists in interdisciplinary projects, resulting in multiple AI product launches.
Built a complete, company-internal AR iOS app from scratch for ML training data collection.
I was the senior scientist on the DermAI product. I built (and patented!) the best detector of melanoma on the market. While there I also built-out explainable-ML features to highlight cancerous regions of a scan to pathologists. Multiple patents obtained here, too!
Longer version:
Implemented, developed, trained and analyzed Tensorflow models for detection and localization of skin, colon and prostate cancer in gigapixel microscope images. Several patents granted.
Led the development of multi-task neural networks that detected melanoma with state-of-the-art accuracy. Abstracts accepted and published by the proceedings of the European Society for Digital Pathology.
Created and operationalized a weakly-supervised neural network that performed data quality control, removing artifacts such as pen ink and air bubbles from microscope slide images.
Generated a biomedical named entity recognition pipeline to exercise on lab information systems.
Mentored junior team members, supervising research projects and code library development.
Created a transfer learning + deep learning + NLP pipeline to understand client questions, and match them to their answers.
Productionalized the model as the backend of a chatbot on the Vanguard website using Domino Data Labs and AWS API Gateway.
Represented web traffic data in a novel, graphical way, enabling me to use Graph Convolutional Networks to identify common web journeys.
This provided actionable insights for improving the website structure and identifying client pain-points.
Developed a Python library to parse millions of highly-unstructured emails into a NoSQL database, and analyzed them using dynamic (time-dependent) topic modeling.
Founded an arXiv journal club for CS/ML/NLP paper discussions and team-building. It has since evolved into a technical seminar, with cross-departmental attendance and 20+ attendees per session.
Dissertation: “Outer Space and Fourier Space: Understanding Foregrounds for Neutral Hydrogen Epoch of Reionization Measurements”
Developed an open source Python library to map atmospheric density using public data from worldwide GPS beacons. It was the first such code to generate all-sky maps efficiently and at any date and time requested.
In order to analyze 100 TB of high-dimensional radio telescope data, I designed a compression pipeline in MySQL, Python and Bash, reducing the data set by a factor of 70. Led the analysis of the compressed dataset, developing new quality-assurance and calibration techniques.
Led a research group of 5 undergraduates, resulting in the development of open source Python libraries. Mentored the group in Fourier Analysis and Digital Signal Processing.
Supervised the cluster used by our collaboration of 60 scientists at institutions across the world (USA, UK, South Africa, Italy).
Presented many public lectures to a wide range of audiences (scientists, non-scientists and children).
Awards:
MPhys Thesis: “X-ray heating from halo collapse during the Epoch of Reionization”
Senior Honors Thesis: “An unbiased survey of gamma-ray burst host galaxies with the Herschel Space Observatory”
Awards:
