Projects on ※

Artemis III Landing Site Digital Terrain Modeling via Shape-from-Shading (SfS)

Fri, 04 Jul 2025 00:00:00 -0400

As part of my work the SETI Institute and for the Artemis Geospatial Data Team, I am developing advanced software pipelines and tooling to produce 1 meter per pixel digital terrain/elevation models (a.k.a. DTM/DEM) using the Ames Stereo Pipeline Shape-from-Shading (SfS) capability for Artemis III landing sites at the South Pole of the Moon. This software consists of Python and SQL via DuckDB for rapid spatial joins and bundle adjust statistic analysis across thousands of LRO NAC images as well as bash pipelines to orchistrate efficient processing on the NASA Ames High-End Computing Capability (HECC) system. The first location I am producing models for is the Mons Mouton Plateau, the largest of the 2024 Artemis III Candidate Landing Regions. We anticipate release of these products in fall 2025.

Fine-tuning Segement Anything (SAM) for Mastcam-Z Mission Operations

Fri, 04 Jul 2025 00:00:00 -0400

In Fall 2023, I fine-tuned the Segment Anything Model (SAM) to produce robust terrain segmentation masks for Mars 2020 Rover imagery in Mastcam-Z and Navcam images to support enhanced image stitching and spectral contrast in a Unity based observation planning tool used by the Mastcam-Z Operations Engineers and Scientists called Viewpoint. The fine-tuned model has been deployed since fall 2023 without requiring retraining to date (July 2025), producing segmentation masks upon image downlink to support daily operations shifts. Fine-tuning used pytorch and lightning on a A-100 80GB gpu with the huggingface implementation of SAM and Weights and Biases for experiment tracking.

Lunar Foundational Model

Fri, 04 Jul 2025 00:00:00 -0400

In early 2025, I was added as a domain specialist scientist on the 4-person first-look science team to an effort to build a Foundational Machine Learning Model for lunar science led by NASA Interagency Implementation and Advanced Concepts (IMPACT) group. The first-look science team has been working to define science requirements and scope for the model, model architecture, software engineering, processing PDS data to ML-ready formats & EDL, and developing benchmark datasets for planetary science applications that will be used to validate and assess model performance for end users.

VIPER Rover VIS Instrument Scientist

Fri, 04 Jul 2025 00:00:00 -0400

My initial role at SETI Institute/NASA Ames was as a member of the VIPER rover science team as an instrument scientist on VIS, the science side of the navigation cameras. In this role I developed software, processes, and pipelines leveraging the Ames Stereo Pipeline, Python, Docker, and existing flight herritage software from JPL called VICAR to produce Panoramas, Mosaics, Stereo and Digital Terrain Models using the Navigation cameras on the mast, the aftcams at the rear of the rover, and the hazcams embedded in each wheel well. I also worked with other engineers on the team to extract camera extrinsics from telemetry and write python based implementations of the forward kinematics of the mast pan tilt unit, masking out rover hardward from imagery, as well as helping assess V&V image products. Other tasks included working on requirements and processes for the Imaging console operations role. A forthcoming publication will include examples of my contributions to this mission.

PSSS 2020

Sat, 16 May 2020 10:16:18 -0400

The Planetary Science Summer School is a highly competitive career development program hosted by the NASA Jet Propulsion Laboratory (JPL) each summer that teaches participants how to develop mission concepts. Many participants have gone on to become Principle Investigators (PIs) for missions and instruments. I particapated remotely, as did eveyone else in the program to develop a Enceladus mission concept we called "Tiger". Specifically I was the attitude control lead, and I was tasked with developing ADC requirements for our spacecraft and instruments from our science objectives which I also actively contributed to. Click the "LPSC 2021 Abstract" link to read more about our mission concept! For more details, please read our paper in the Ocean Worlds focus issue for the AAS Planetary Science Journal "PSJ Paper"

Recently, David W. Brown wrote a fantastic article about the program in The New Yorker titled: "How to Plan a Space Mission" .

To learn more about the program, checkout this article from JPL about our session of the program and this video from NASA JPL about the program from 2014 to learn more about the final week of the program:

OpenSource

Sat, 16 May 2020 10:12:38 -0400

SpiceyPy

Sat, 16 May 2020 10:02:26 -0400

SpiceyPy

SpiceyPy is an open source Python wrapper for the SPICE Toolkit that I started developing in 2014. SpiceyPy is now used extensively by scientists, engineers, and hobbyists for their work. SpiceyPy is being used now to plan trajectories for active Spacecraft, for scientific research, and to produce planetarium shows (see below). Over the years, I have presented multiple talks about SpiceyPy, primarily at the biennial Planetary Data Workshop, so many of the technical details can be found in those works. More recently, SpiceyPy was peer-reviewed and published in the Journal of Open Source Software (JOSS). More about SpiceyPy, including links to the github repository can be found above.

Worlds Beyond Earth

I was also in the credits for the recent American Natural History Museum planetarium show "Worlds Beyond Earth" for writing SpiceyPy, which was used to help generate the visualizations of trajectories of planets and spacecraft.

FDL 2019

Thu, 16 May 2019 10:10:35 -0400

The Frontier Development Lab is a highly competitive research ~10 week internship experience held at SETI in Mountain View California each summer. I was placed on the Floods team, where our research goal was to apply machine learning to predict flood events from USGS river gage data, using the Google Cloud to accelerate the science. I had a awesome time in California as well and our work resulted in multiple NeurIPS conference workshop papers. To learn more about our work, check out the video link above and the video about our work produced by Google Cloud below!

Python Hapke Model

Tue, 16 Apr 2019 10:15:53 -0400

In 2019 I implemented a Hapke photometric model using Python and Numba (there are many flavors) from the 2012 textbook by Hapke for a Space Weathering class project. I used Numba to gain massive speed improvements without using complications like cython.

CRISM Hyperspectral Mapping

Fri, 16 Nov 2018 10:13:30 -0400

In 2018 I started working with the science operations team for the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) instrument. My role is to help Pythonize a pipeline implemented in IDL for a new mapping product from one of the CRISM observation modes (HSP). My work will help make this mapping product available to the scientific community to use in studying fine spectral properties of the surface of Mars at a moderate spatial resolutions. So far, some of my work from this effort is already publicly available in the form of code contributions to QGIS.

Deep Layer Mapper

Fri, 16 Nov 2018 10:12:07 -0400

For my Ph.D. research, as I explain further in my Arabia Terra project, I am mapping traces of the layered deposits in Arabia Terra in order to measure how thick each layer is and to investigate their structure from their geometry. I am mapping the layers in HiRISE images that I am also producing Digital Elevation Models (DEMs) which cover relatively large areas of Mars at 0.25 cm per pixel resolution. Mapping the deposits by hand is tedious busy work, and is distracting from the science I want to accomplish. To accelerate the mapping of these layered deposits, in 2018 I developed a training data set from layers I had already mapped and trained a U-Net architecture Convolutional Neural Network (CNN) to automate the process. Specifically, the CNN was trained on 1 meter per pixel normalized HiRISE images, so it would not rely on the elevation models to map my layers, and was provided many true negative examples of other similar surface features such as linear dunes which were not desireable to map. Traditional computer vision techniques need to be more deeply investigated, but as there is a large number of criteria that I use when mapping the layers that would be difficult to quantify, using a neural network allows me to quantify my expert knowledge implicitly from the training data. I presented an abstract and talk about this work at the 2018 NASA Goddard AI Workshop.

Regional Stratigraphy of Arabia Terra

Fri, 16 Sep 2016 10:11:08 -0400

Intro

I started my Ph.D. at Johns Hopkins University in the fall of 2016. This page will expand to a few paragraphs to explain my research project so please stay tuned for updates. Given the amount of work I've done, it will take me a while to update this page so below I'll just preview the section titles to hint at what will eventually be here. Until then please read my recent JGR Planets Paper (link is in my publications page).

some cool figures

Layered Deposits Big Picture

Layered deposits have been found across the surface of Mars since we started taking high-resolution satellite images of the surface. Layered rocks are essential to planetary scientists because the layers can preserve a record of the climate that describes the history of the planet. The presence of layered rocks alone tells us that some re-occurring process produced the different layers, and these processes can involve liquid water. Knowing the precise details about the activity of liquid water on Mars are one of the scientific community's biggest unanswered questions. By investigating layered rocks, in a variety of locations and interpreting their geologic record, we can piece together the larger story of water on Mars.

Layered Deposits in Arabia Terra

Geography on Mars is a bit different from the Earth due to the lack of plate tectonics and oceans. A broad region near the equator and prime meridian called Arabia Terra contains many old impact craters containing layered deposits within their interiors. If you have read the book or seen the movie The Martian by Andy Weir, Arabia Terra is the region the character Mark Watney drives across (no spoilers). These layered deposits, in particular, are interesting because they may have formed by lakes or playas within the craters fed by groundwater. I also officially named a few of these craters with The Gazetteer of Planetary Nomenclature through my first paper.

Potential climate signals rocks record

On the Earth, layered rocks are formed by the deposition of sediment to form sedimentary rocks. As these layers accumulate, they each say something about the environment in which they were created, and in particular, the time they were formed. Layered rocks serve as a literal record of geologic events, much like tree rings record a tree's history. These rock records describe the history of the planet they were formed on. This principle underpins geology and is how we have learned the geologic history of the Earth. In particular, understanding how the layers were formed is critical to know what the record is telling you. So if we can figure out how these layered deposits were formed, we can figure out what the climate of Mars was like when they were created and how it may have changed through space and time.

HiRISE is amazing

To study these layered rocks, I need to get close. Unfortunately, the rovers on Mars are busy elsewhere. Instead of looking at these rocks from the surface (in-situ being the term), I use satellite images from a camera called HiRISE (High Resolution Imaging Science Experiment). HiRISE is onboard the Mars Reconnnasaince Orbiter spacecraft and images the surface at about a quarter meter per pixel resolution. This resolution is comparable or better than what is commercially available on the Earth, at least until fairly recently. Thanks to this high resolution, I can clearly see these layered deposits in Arabia Terra. Although to properly study these deposits, a single picture is not enough to answer how the layered deposits were formed.

Semi-Automated Stereo Photometry with the Ames Stereo Pipeline

However, suppose I can get two carefully planned images as a stereo pair. In that case, I can make a digital elevation model (DEM) of the surface. This process is called stereophotogrammetry or computer stereo vision. It can produce an elevation dataset at a similar high resolution to the images used. However, HiRISE images are large, and stereo processing can take lots of time and manual effort. To solve this problem, I used open-source software provided by NASA called the Ames Stereo Pipeline (ASP). ASP automates much of the tedious work required to make a DEM. Through my Ph.D. I have developed a toolset that further automates ASP to make HiRISE DEMs called ASAP (Ames Stereo Automated Pipeline). I describe ASAP in a separate section of my site, and it is hosted on my GitHub. In the end, I can produce high-quality DEMs of the layered deposits in a reproducible manner.

Importance of elevation data

Elevation data is critical for studying these layered deposits as it is a massive information dimensionality increase from a single image. A lot of geology is possible from a single image. However, quantitative answers to questions about layer thickness and directionality (orientation) are tough without elevation data. From the elevation data, we can visualize and measure the geometry of the layered deposits. We can determine how these layered deposits were formed from this geometry information by comparing the geometry with deposits studied on the Earth. We also gain a record of each layer's geometry and can see how these layers changed through time. The changing properties through time describe how each layer's formation varied and can, in turn, tell us how climate conditions on Mars changed.

Deep Layer Net (see other project for more)

First Paper tl;dr

CIRS

Sun, 16 Jun 2013 10:14:27 -0400

In 2013 I was an intern at NASA Goddard Space Flight Center (GSFC) working with Dr. Conor Nixon on the Composite Infrared Spectrometer (CIRS) instrument science team. During my internship I used Python to produce plots of the spatial and latitudinal coverage of science observations of the moon Titan from the instrument for a number of observation modes. Maps and plots I produced were used to prioritize missing observations of Titan to be made by the instrument prior to the end of the Cassini mission in 2017. Recently, a paper including me as a co-author was published in the ApJS journal (pdf linked above). The internship application was highly competitive.

Enceladus

Mon, 16 May 2011 10:14:39 -0400

As an undergraduate student at The University of Virginia I worked with Dr. Anne Verbiscer to process and calibrate images of Enceladus from the ISS NAC/WAC cameras on the Cassini Huygens spacecraft. In addition to calibrating the data, I needed to subset the pixels from the images by the geologic unit the pixel belonged to. The goal of the project was to ultimately solve the inverse problem of Hapke photometric parameters corresponding to the units to better understand the ices and the surface properties of Enceladus. An additional requirement was that the image could not be interpolated in any way, so instead of map-projecting the images to a common basemap, I implemented the ISS camera model in python to project the geologic map into the camera image plane. This project resulted in a number of abstract publications and will hopeful lead to something peer reviewed someday.