REHS Project Descriptions

This position requires a successful student candidate to possess excellent skills in photography and videography. The student intern must be able to create a photo gallery that captures the essence of the summer programs at SDSC. Photos taken shipment of high-quality and should not be blurry, out of focus, ridiculing or in part taste. In summary, all photos should be professional. They student intern should have excellent videography skills, being able to create numerous short videos per week for website and social media purposes. The student’s photos and videos will be shared on the program website, LinkedIn and Instagram.

It is important that Easton intern be able to take photos and videos during lectures and project work time without being obtrusive. Students will receive publish credit for their work when applicable.

If the student candidate is granted an interview, the student candidates will be asked to share a recent photo gallery and two short videos (Instagram Reels) to help interviewers assess the candidates skill level. Images created by a successful intern may also be shared with

UCSD Communications as part of our community outreach efforts.

[Please Note: this internship will run from Monday, June 29 – Friday, July 31, 2026.]

The successful student intern will act as a program assistant to our 2 undergraduate teaching assistants who will be helping with our UCSD Advanced Research Experience (ARE) program this summer. Duties could include assisting the weekly professor with copying request, student participant check in each morning, technical issues, participant form review and other tasks as they arise.

This position is a morning– only position. The selected intern will be permitted to listen to the morning ARE lectures as time permits.

The COMPLECS Training Program is an applied learning experience designed to help Advanced cyberinfrastructure users aquire the skills and knowledge they need to efficiently accomplish their compute and data intensive research and effectively using High Performance Computers. COMPLECS focuses on building technical confidence through hands-on training, and exposure to the tools and workflows used in modern scientific computing environments.

The primary goal of this research project is to develop training material for foundational knowledge in Advanced Cyber infrastructure (CI) and high-performance computing (HPC). Participants will be introduced to Advanced-CI topics and HPC computing by review current training material and be asked to review existing training materials, COMPLECS training content– including slides, documentation, and hands-on exercises– and contribute to the development of new training modules and practical learning activities that support the program’s mission.

During the course of this project, you will be introduced to HPC and learn how to:

• work with advance CI infrastructure and tools;
• run compute jobs on a supercomputer like Expanse via the Slurm workload manager; and
• develop professional training material using Github, Powerpoint, Linux

The Science Gateways Center of Excellence (SGX3) is looking for talented communicators to explain cyber infrastructures and the research they are conducting to the community. To do this, the communicator will conduct research about these cyber infrastructures – which we call science gateways, understand each science gateways’ research applications to societal challenges, and document how they are solving research challenges or instructing the next generation. Tasks will include conducting your own research about multiple science gateway projects, writing down findings in a digital document to be shared with your mentor, and writing a story about the science gateway. During the summer you will publish several stories about these science gateways with support of your mentor. You will also work with the SGX3 team to publish these stories in monthly newsletters and on the sciencegateways.org website. The successful student candidate(s) will meet regularly with their mentor to review tasks and provide suggestions and feedback. The student candidate(s) will also have access to other communication channels to talk to other members of the SGX3 team via Slack and email.

This project involves developing and testing AI based software and Jupyter Notebooks to run on NSF funded resources including the Expanse high-performance (HPC) system [1,2] and the NSF funded ICICLE AI project [3]. Expanse is SDSC’s largest supercomputer. The result of a $10M National Science Foundation (NSF) award, Expanse delivers over 5.2 peak PetaFlOps of computing power to scientists, engineers, and researchers all around the world [2]. Expanse provides three kinds of HPC/CI resources: General Computing Nodes (CPU), NVIDIA GPU Nodes (A100, H100), and the petascale Luster filesystem. Thousands of users have accessed these high-performance computing (HPC) resources via traditional runs from the command line and using batch queuing systems.

The National Science Foundation funded AI institute for Intelligent Cyberinfrastructure with Computational Learning in the Environment (ICICLE) [3] is building the next generation of Cyberinfrastructure to render Artificial Intelligence (AI) more accessible to everyone and drive its further democratization in the larger society. ICICLE is developing intelligent cyberinfrastructure with transparent and high-performance execution on diverse and heterogeneous environments. It will advance plug-and-play AI that is easy to use by scientists across a wide range of domains, promoting the democratization of AI. The list of currently available ICICLE software can be found in the Training Catalog [4]: https://icicle-ai.github.io/training-catalog/

Scientists using HPC Systems often work with interactive HPC tools such as Jupyter notebooks to implement computational and data analysis functions and workflows [5]. Jupyter notebooks are web applications that allow you to create and share documents that contain live code, equations, visualizations and narrative text. These notebooks are part of a general trend in research computing away from command-line style interfaces and towards browser-based and graphical interfaces. Jupyter notebooks are especially useful for interactivity: the development, testing, and exploration of data sets or as an instructional resource [6][7]. Users working interactively expect a timely response, both for initial application startup and during the course of a session.

The goals of this research project will be to:

• Learn basic parallel computing concepts;
• Test and develop parallel Jupyter Notebooks that run on Expanse;
• Test software components developed by the ICICLE project; and
• Learn AI/ML/DL concepts.

The research components will be to:

• Contribute to the body of knowledge needed for hosting live, dynamic, interactive services that interface to HPC systems; and
• To develop interactive notebooks that will run using ICICLE software components on an HPC system.

Students will also have the opportunity to publish their results on an open share site such as arXiv.org [8], and have their work featured on project pages [9] and the ICICLE Education Site [10].

Large Language Models (LLMs) are now widely used for a number of natural language processing tasks including content generation and knowledge retrieval, information synthesis and summarization, and software development and workflow automation. Unfortunately, however, many of the most powerful LLMs available today are commercial products only accessible to end users via either a web-based interface or an application programming interface (API). Moreover, in most cases the data provided by end users to these LLMs may be retained by the company for their own future use. This presents a challenge to the use of these LLMs when evaluating and analyzing protected data and comparing them against open source models that can be run locally in a more secure environment.

The fundamental aim of this research project is to develop a secure pipeline to evaluate and analyze protected data with LLMs. The protected dataset of interest here is a subset of the SDSC’s internal user support ticket data, which tracks email-based interactions and exchanges between the SDSC user community and the User Services Group. Ticket issues include, but are not limited to, managing user accounts, answering general user inquires, debugging technical problems reported by users, and making best practice recommendations on how users can achieve high-performance when running their scientific workloads. However, these tickets almost always contain personally identifiable information (PII), and sometimes even login credentials. Therefore, in order to leverage the use of commercial LLMs in the evaluation and analysis of this type of ticket data, we need to first discover and then remove and/or mask all protected data before we can submit prompts and obtain responses for further analysis. 

During the course of this project, you will learn how to:

• parse text-based data in Python;
• build and label a non-trivial natural language dataset;
• securely run LLMs on the Expense supercomputer at SDSC; and
• make Python-based API calls to commercial LLMs

The EarthCube Office strives to assist geoscientists and the wider research community through tools, methods, standards, architectures, and community connections. Our group seeks high school students that are interested in writing pieces related to EarthCube projects from start to finish. As a result of this internship, students will come away with an ability to reduce technical jargon and other barriers for the creation of easy-to-understand written pieces. Students will have the opportunity to choose a geoscience domain that is of interest to them and create both written and video pieces regarding that topic.

The Neuroscience Gateway (NSG – https://www.nsgportal.org) provides access to supercomputing resources for computational and cognitive neuroscientists. Through a simple web-based portal, the NSG provides a user-friendly environment for uploading models or data, specifying supercomputing job parameters, querying running job status, receiving job completion notices, and storing and retrieving output data. The NSG distributes user jobs to appropriate supercomputing resources. REHS students will learn about supercomputing and NSG during the initial period of the internship. Followed by that REHS students will be involved in either computational modeling of neurons using the NEURON software or processing of Electroencephalography (EEG) data using the EEGLAB software. In computational modeling scientists build models of neurons or network of neurons that perform various functions in our brain. NEURON is a widely used software for such modeling of neurons. Students will learn about a modeling project as part of their internship. EEG project involves recording of brain’s electrical activities tied to some function a subject is performing, and processing the data using software (such as the EEGLAB) and computing resources to see connection between brain signals and the function. Students will record and process EEG data as a part of their internship.

We are interested in students who have some background and interest in programming and scripting and are interested in neuroscience, computer science, data processing and modeling. Some experience with programming languages and exposure to Linux systems are preferred.

Our group seeks high school students who are interested in journalism with a focus on learning more about effectively communicating the results and societal benefit of science, engineering, and technological research projects. Candidates should have a strong desire to learn more about how to communicate technological research — with a keen eye for making the work both engaging and easily comprehended by a general audience. At the same time, students should be mindful of not diminishing the technical work of the research team in any way. Sound challenging?

As a result of this internship, students will come away with an ability to reduce technical jargon and other barriers for the creation of easy-to-understand written pieces. Students will have the opportunity to choose a science, engineering, and/or technological domain that is of interest to them and create written pieces regarding that topic as related to supercomputing.

Students will be assisted with both the science writing process as well as the final editing and completion of their articles with accompanying images, captions, and credits. 

In this project, we will explore how modern machine learning approaches can be employed in computational chemistry. As a background, the work in our lab encompasses computer simulations based on quantum chemistry and molecular dynamics. We develop numerical models and software for atomistic simulations of condensed phase systems and apply these to solve problems in the chemical sciences, ranging from atmospheric chemistry to biochemistry and drug design. Our focus has been on multiscale methods that couple computationally demanding quantum descriptions of the electronic structure with computationally simpler classical interaction potentials in so-called QM/MM methods. This enables numerical simulations of realistic models that are intractable with pure QM methods. The importance of such methods has been recognized with the award of the 2013 Nobel Prize in chemistry for “the development of multiscale models of complex chemical systems”. While machine learning approaches have a long history in computational chemistry, recent advances in data storage and compute capabilities have ushered in a new era of data-driven approaches, which we seek to explore and combine with existing QM/MM models. We are involved in the development of several software packages that are used by many research groups in academia and industry. This includes the quantum chemistry software packages ADF and QUICK and the molecular dynamics simulations package AMBER. Several other major codes that are deployed on major national computer resources are also an integral part of our work.

In this REHS project, the students will explore machine learning approaches that may improve the speed or enhance the accuracy of current state-of-the-art QM/MM simulation approaches. We will jointly work through online course material that introduces machine learning concepts from simple classifications to deep learning, culminating in applications that predict chemical properties. We will make use of Jupyter notebooks, the Python programming language, and widely used machine learning software frameworks. We will explore the performance of the software on CPU and GPU hardware including the Expanse supercomputer at SDSC. Time permitting, we will generate training data using quantum chemistry software and develop models that can predict new data without having to run the expensive QM simulations in the first place. The interns will thus have the possibility to acquire a wide range of skills in data science, computational chemistry, and high-performance computing. As an essential part of the internship the students will learn how to document their research, prepare research reports and present their results to their peers – skills that are very important for a successful research and engineering career.

Introduction to Artificial Intelligence (AI) and ML (Machine Learning) on the Edge

The Introduction to AI and ML on the Edge is an interactive, project-based program where students collaborate in teams of three* to explore and apply AI and ML techniques using low power embedded systems and AI accelerators. Participants will receive specialized AI acceleration hardware running on embedded Linux to work throughout the program. Moreover, students will have access to the UC San Diego Supercomputer Center for advanced AI models’ data preparation, training, and benchmarking.

In the initial phase of the program, students will learn how to create and deploy AI and ML models tailored to the provided hardware. Following this, students will work on real world problems such as autonomous vehicles: AI-Powered Autonomous Garden Ecosystem (AI-AGE) Project. 

The program will focus on Artificial Neural Networks, Computer Vision, and, depending on student progress, may also explore Large Vision Models (LVMs).

The Open Science Chain (OSC – http://www.opensciencechain.org) project uses blockchain technologies to protect the integrity and provenance of research artifacts. The project is looking for summer interns who are interested in working with the OSC project. Interns will have the opportunity to engage in various tasks, including developing blockchain applications, revamping the web portal, or assisting in unit test development, depending on their interests and expertise levels.  Programming experience (Go, Java etc) and familiarity with Linux and container technology are preferred. This is a completely virtual internship. 

The MLPerf Inference Benchmarks are designed to measure and evaluate the performance of artificial intelligence (AI) and machine learning (ML) inference workloads on different computer hardware and software platforms. The benchmarks include problems in computer vision, natural language processing, and generative AI.

The fundamental aim of this research project is to develop a pipeline to automate the process of building, testing, deploying, and running a subset of the MLPerf Inference Benchmarks on the Expanse supercomputer at SDSC.

During the course of this project, you will learn how to:

• set up and run inference jobs on Expanse via the Slurm workload manager

• evaluate the performance of AI/ML inference workloads

Have you ever used ChatGPT or other AI assistants? Have you wondered how they work? In this project, you’ll help build an AI-powered chatbot that assists researchers and students using the National Research Platform (NRP) – a powerful computing system that connects research institutions across the country. Think of it as creating a helpful assistant that can answer questions about using supercomputers and cloud computing tools!

The chatbot you’ll help create will use Large Language Models (LLMs)– the same technology behind ChatGPT – to understand questions and provide helpful answers about Kubernetes (a system for managing computer applications) and High-Performance Computing (HPC). Instead of researchers having to search through complicated documentation, they can simply ask the chatbot questions like “How do I submit a job?” or “Why is my program running slowly?” and get instant, helpful responses.

In this internship, you will learn how to:

• work with AI language models (LLMs) to create conversational interfaces
• build a chatbot that can answer questions about Kubernetes and HPC systems
• create a web interface where users can chat with the AI assistant
• train and improve the chatbot by providing it with knowledge about computing systems
• integrate the chatbot with the National Research Platform infrastructure

This project is perfect for students interested in artificial intelligence, chatbots, web development, or helping others solve technical problems. You don’t need to be an AI expert – we’ll teach you what you need to know! You’ll work with Python, learn about how modern AI assistants work, and get hands-on experience building something that real researchers will use.