Martin Chobanyan

Hi, you found my webpage!

I am a machine learning scientist who loves to explore new domains and use machine learning to solve difficult problems.

You can find more about my work and interests below.

Experience

Senior Machine Learning Scientist

Jul 2021 - Present

At SmileDirectClub, I explored the reconstruction, positioning, and generative modeling of 3D teeth using multi-view customer images.

In particular, I was one of the key developers of the 2D-to-3D dental arch reconstruction model for SmileMaker Platform, a real-time app which delivers personalized treatment plans for thousands of daily customers using completely self-guided smartphone imaging.

We have released a patent for the positioning of 3D model teeth using information across multi-view, 2D images.

Data Scientist

Jan 2018 - Jul 2021

CCRi is a geospatial analytics company where I worked on various research projects involving maritime activity detection and vessel forcasting from large scale AIS trajectory data.

I presented my work in a speaker role at the 2019 Tom Tom Applied Machine Learning Conference in a session named Points to Videos: Extracting Behavioral Information from Maritime Tracks Using Convolutional Neural Networks.

Bachelors with Distinction

Aug 2014 - Dec 2017

I graduated from University of Virginia in 2017 with a double major in Computer Science and Statistics with a focus on artificial intelligence.

Personal Projects

Visualizing Change in Neural Networks

An experiment to assess the change in feature visualizations of a neural network as it is fine-tuned to a new task

Emotion Detection with Transfer Learning

Creating an emotion detector using pre-trained computer vision models, transfer learning, and a nifty way to create a custom dataset using Google Images

Competitions

CHAMPS Molecular Prediction (Top 5% Finish)

Predicting molecular properties using graph convolutional neural networks and chemical descriptors

HPA Single Cell Classification (Top 6% Finish)

Segmenting the locations of target proteins within cell microscopy images using a Transformer-based architecture

LANL Earthquake Prediction

Predicting the next seismic event in simulated earthquake data using an ensemble of deep learning techniques