Science & Engineering Apprenticeship

Summer students in 2019

Become an Apprentice

Applications for summer 2020 are no longer being accepted. Applications for 2021 will be accepted in January.
  • - Must be a graduating high school senior who will attend a senior college in Fall 2020
  • - Must be available to work throughout the summer appointment period (Unpaid time off is allowed for attendance of freshman college orientation, otherwise there are no exceptions to the policy).
  • - Must have both applied and been admitted to The University of Texas at Austin
  • - Must be a U.S. Citizen

About the Program

The Science & Engineering Apprenticeship Program is a competitive program for graduating high school seniors who plan to attend college in the fall semester following graduation and who have both applied and been admitted to The University of Texas at Austin. The program exposes the selected students to laboratory research and development and gives them an idea of what scientists and engineers do. Local area high schools are notified of the program, and students who have an interest in electrical, mechanical and aerospace engineering, physics, mathematics, and computer science are invited to apply. U.S. citizenship is required.
Each student is supervised by a research scientist or engineer and given a project they can complete during the apprenticeship. In addition to these projects, the program offers students opportunities to learn more about the different types of research done at ARL:UT. These activities include presentations from ARL:UT staff, a tour of ARL:UT's Lake Travis Test Station, and a chance for the students to showcase their research through technical reports and poster presentations.
The Program began as a part of the Department of Defense Science and Engineering Apprenticeship Program for High School Students and encourages students to pursue careers in the science and engineering disciplines, particularly in areas related to the needs of the U.S. Department of Defense. ARL:UT accomplishes this goal by carefully assigning each student to a research project that can be completed during the summer. When the program started the summer of 1982, nine students from five local high schools participated. Now, over 561 students have taken part in the program, and most have gone on to major in science or engineering in colleges throughout the United States. Many participants return to ARL:UT in student and research positions and stay on to contribute for several years.

2019 Student Projects

Due to the process involved in obtaining sediment cores, the measurements of physical cores' geoacoustic properties may not be accurate, as the collection process will disturb the sediment sample. In an attempt to resolve this problem, an acoustic coring system was developed that is capable of taking in situ measurements of shear and compressional wave speed and attenuation. Field testing of this system revealed that, due to the signal to noise ratio of the shear waves, the signal was unattainable in the presence of noise from the core's penetration into the seabed. To analyze and address this issue, a set of lab experiments were conducted in order to compare shear wave speed and attenuation measurements at variable depths when static versus while moving. Shear wave measurements were acquired using the aforementioned acoustic coring system, which creates and detects the shear waves using bender elements mounted in flat blades. The measurements were made in Hydrite Flat DS kaolinite mud, a porous sediment whose geoacoustic properties have been previously measured. This paper describes the setup of the experiments, explains how the data from the acoustic coring system were obtained and analyzed, and evaluates the impact of interference due to sediment penetration on shear wave velocity and attenuation measurements.
Westlake High School, Austin, Texas, Environmental Sciences Laboratory
While machine learning has become a pivotal tool for image classification, current methods rely on a substantial amount of labeled data in order to ensure robustness. Here, we examined few-shot learning algorithms to identify aspects of these algorithms that can lead to more robust learners. These algorithms are designed to perform well on new classes when trained with limited (one or five) labeled examples from each new class. Specifically, we trained a Convolutional Neural Network using state-of-the-art few-shot methods from Chen et al. [A Closer Look at Few-Shot Classification (2019)] to identify characters from human scripts in images and investigated their efficacy in classifying out-of-domain symbols. These methods include both traditional transfer learning approaches as well as meta-learning algorithms which use meta data to choose parameters. We trained a four layer CNN on the Omniglot dataset, which consists of images of characters from many human scripts. Then, we tested our model on script-like images (e.g., zodiac symbols) which are similar to the Omniglot characters, as well as on images from a drastically different domain (e.g., silhouettes of objects). Surprisingly, we show that these few-shot algorithms work effectively for both small domain shifts (script-like) and large domain shifts (object-like) suggesting that accuracy can be maintained fairly well for varying kinds of domain shifts. Ongoing experiments are being conducted to better understand the robustness of the models.
Westwood High School, Austin, Texas, Signal and Information Sciences Laboratory
ARL:UT has identified a need for portable battery powered timing modules that can hold sub-millisecond accuracy for a duration of several weeks. While millisecond-level timing is not a challenge with GNSS timing receivers this system can not rely on external signals to maintain accuracy. Several Oven-Controlled Crystal Oscillators (OCXOs) were considered due to their low prices and compact sizes; however, the Chip Scale Atomic Clock (CSAC) was eventually selected to solve the issue of OCXO's comparatively high drift rates and power requirements. The Microsemi CSAC best fit these requirements with its low drift rates and 120 milliwatt power consumption. An Arduino Due was selected as the timing module microcontroller. Using the Arduino microcontroller, an interrupt based timing program is being developed to register the 1PPS output of the CSAC. The system will detect a pulse, immediately set the corresponding microsecond timestamp to the detected pulse, and output the timestamp to an interface port. For prototyping, a function generator and oscilloscope will serve together as a surrogate pulse output, allowing software to be tested without risking damage to the CSAC. After finalizing the software, the CSAC and the Arduino Due will be integrated through the CSA C's proprietary development board using soldered SMA adapter cables. After successful integration, the timing module will undergo longer duration testing to quantify the module timing accuracy.
St. Stephen's Episcopal School, Austin, Texas, Space and Geophysics Laboratory