Explore Projects

The proposal titled "Event Classification With Masked Transformer Autoencoders" aims to enhance particle jet tagging by developing a Lorentz-equivariant Particle Transformer (Lorentz ParT) that adheres to the symmetries of special relativity. This hybrid architecture combines Lorentz Geometric Algebra Transformers (L-GATr) with Particle Transformers (ParT) to compute physical invariants, such as invariant mass, while tracking geometric orientations between particles. A key innovation is the introduction of Gated Particle Attention, which addresses residual bottlenecks in standard transformers by dynamically weighting contextual information against original features. The model is designed for a dual-head functionality, performing both self-supervised classification across ten unique jet classes and continuous mass regression using the JetClass dataset. To optimize learning, a Masked Autoencoder (MAE) approach is utilized to understand underlying jet symmetries before label-based training begins. Preliminary results indicate that the gated prototype can achieve a lower eta loss than original methodologies, although it currently faces challenges with "hallucinating noise" in reconstruction maps due to increased parameter counts. The proposed 12-week execution plan focuses on establishing mass regression, mitigating ROC curve issues, and validating the architecture against a scratch model to ensure stable performance

Summary: The project aims to extend istSOS to support the dataArray extension, which is commonly used in SensorThings for managing Observation entities. Currently, istSOS lacks support for dataArray, limiting its compatibility with SensorThings. By incorporating dataArray support, istSOS will become more versatile and user-friendly, allowing for efficient retrieval and creation of Observation entities. This extension is expected to enhance istSOS's performance by reducing the size of requests and responses, ultimately improving the user experience. The project's expected outcome is the successful integration of dataArray support into istSOS.

JPEG XL is an image format that allows for lossy and lossless compression, and is generally better than JPEG at this. LibreOffice currently supports many image file formats, but JPEG XL is not one of them. This project would add such support. The potential for storage savings that support for this image format would bring will mean a better experience for experienced and newer LibreOffice users alike. In addition, web browsers are increasingly supporting JPEG XL (it is available in nightly Firefox builds), so this project will also help LibreOffice users by improving compatibility.

WasmEdge is a WebAssembly runtime that supports both interpreter and ahead-of-time modes. For proxy-wasm support, WasmEdge only provides the interpreter mode currently. Such as the other runtimes, WasmEdge should be able to support the AOT mode for better performance. In this mentorship, the mentees will help the WasmEdge project to complete the AOT mode in proxy-wasm proposal and write the docs for examples of running with proxy-wasm. - Expected Outcome: - Modify the Bazel file to include the LLVM dependency. - Modify the code to support running WASM in AOT mode. - Add the documentation of proxy-wasm in the WasmEdge docs repo.

Today, with Ubuntu increasingly adopting Snap packages, Uyuni currently lacks support for managing these applications. This project proposes adding native Snap support to Uyuni, including install, update, remove, and channel management. By extending Uyuni’s capabilities to cover Snap, this work fills a key gap and ensures continued relevance for Ubuntu system management. I explored how Uyuni manages CVE audits and package upgrades for Ubuntu systems. Initially, Uyuni connects to client machines via SSH to install and register the Salt Minion. Once set up, all communication occurs through Salt’s remote execution system, eliminating the need for SSH. Salt gathers package data using grains, pillars, and custom modules, which Uyuni then compares to its centralized SUSE Manager database. For CVE auditing, Uyuni fetches security metadata (e.g., CVEs, OVAL data) from upstream sources like Ubuntu Security Notices and stores it in a PostgreSQL database. When Salt reports installed packages, Uyuni analyzes the data and generates patch or upgrade recommendations. APT repositories are mirrored into structured software channels, allowing clients to receive updates without direct internet access. Building on this workflow, I propose extending Uyuni to support Snap packages. While Snap lacks structured CVE metadata, we can implement a version-based upgrade system: Uyuni retrieves the latest Snap versions from the Snap Store and compares them to client versions reported via Salt, suggesting upgrades when newer versions exist. To support airgapped environments, we can pre-download .snap and .assert files on a connected system and distribute them through Uyuni’s software channels. Additionally, to manage storage and stability, Uyuni admins can choose which Snap channels (stable, beta, etc.) to mirror—defaulting to stable. Deliverables: Support Snap auditing, lifecycle management, airgapped environments, testing/validation, and user documentation in Uyuni.

The torch ecosystem is rapidly growing in R, but it still lacks support for several key computer vision capabilities available in Python's torchvision package. This proposal aims to close the gap by implementing a wide range of torchvision datasets and models directly in R. Specifically, the project will: Implement all torchvision datasets. Add support for 5 key computer vision tasks: Object Detection, Instance Segmentation, Keypoint Detection, Semantic Segmentation, and Quantized Models. Provide native R implementations that don't rely on Python bindings, making it easier to install and use. Deliverables include tested R implementations of datasets and models, documented APIs, and example use cases. This will make R a more viable option for computer vision research and production.

<p>JPF is a model checking tool for java applications. JPF-core is the core structure of JPF. The build for jpf-core has been moved from ant (upto java 8 support version)to gradle. The current JPF-core version doesn’t have java 11 support i.e jpf-core is not portable for java 11. Jpf extensions have not moved from ant to gradle yet. This is because of the potential breaking changes from its previous versions. One such major breaking change in java 11 is “bootstrap methods”. The goal of this project is to fix gradle support for java 11, to update the extension template,provide the widely used jpf-extensions with gradle support.</p>

<p>JPF is a model checking tool for Java applications. It is a Virtual Machine (VM) for Java bytecode which executes the system under test (SUT). The JPF source is partially compatible with Java 11. The problem is to support new features of Java 11 like the bootstrap methods. They are used for things as common as string concatenation ( String + String, String + int, String + double, String + Byte, etc ). As of now, a few specialized cases are supported, but there are still few cases that fail with Java 11. There are also some internal APIs from Java 11 that no longer exist in Java 12. Therefore the aim of this project is to support the general case of bootstrap methods and update the code to support java 12.</p>

SQLancer is a tool that automatically tests DBMS in order to find bugs in their implementations. As of right now, SQLancer supports many great database systems, however, some of the systems are outdated including one of the most frequently used databases: PostgreSQL. SQLancer supports Postgre version 12 which became end-of-life in Nov 2024 and many versions along with critical changes have been made since then. This project aims to advance SQLancer to support the latest version of Postgres (version 17) by removing necessary deprecations as well as adding new features, by the end of this project, SQLancer will support additional features such as JSON data types/functions, procedural extensions, advance error handling, and hopefully reiterate the upgrade to other databases

<p>Currently, the Google-Backend for GVfs supports only a subset of operations allowed by Google-Drive on the web-interface. The major problem with supporting all the operations is because of the difference in how POSIX systems handle files whereas how a database-backed system like Google-Drive handles files. This difference results into limitations as to what operations can be performed with the current libgdata API. Since, each file’s identifier is its ID, i.e. “name” equals ID, we have to specifically use “display-name” for storing a file’s title, which is what is shown by nautilus.</p> <p>Simply copying/moving files from one folder to other folder results into an error “Operation not supported” so as to preserve the file’s title. Copying/Moving is one of the fundamental operations that should be possible on a file. My ultimate goal with this proposal is to add support for this necessity and make the Google-Drive backend more usable.</p>

APNG (Animated PNG) is an animated format backwards compatible with PNG. It is supported by all the major browsers and has benefits over GIF such as partial transparency and support for more colors. The goal of this project is to integrate APNG support in LibreOffice, which had been requested for quite some time.

<p>Currently, there are no supports for S3 and GCS on Windows. It leads to two problems:</p> <ul> <li>The limitation of storage capacity and inconvenient of traditional hardware in contrast with cloud base storage.</li> <li>The overhead when using TensorFlow cross-platform as Linux and macOS have already been supported.</li> </ul> <p>Furthermore, according to this <a href="https://github.com/tensorflow/tensorflow/issues/19297" target="_blank">discussion</a>, @mihaimaruseac said that with the implementation of modular filesystem C API, the current S3, GCS and Hadoop support on Linux and MacOS will be moved out of TensorFlow, into SIG IO. New plugins will be necessary to ensure the continuous data accessibility.</p> <p>Therefore, this project aims to provide complete S3 and GCS support for Windows <strong>and</strong> write new plugins for S3,GCS and Hadoop for future use when the modular filesystem C API is ready.</p>

This project extends GNOME Crosswords' existing rendering facility to support printable puzzles with clues, and integrates printing support with crossword-specific options. It also has stretch goals, including broader puzzle support and an `ipuz2pdf` export utility.

<p>MIPv6 is the base standard from the IETF for IP based mobility management in wireless networks. The next generation mobility management schemes are mainly following the protocol stack of MIPv6. Until today ns-3 – which aims to replace the aging ns2 and its associated old-featured simulators – does not have even in its latest version a MIPv6 module that is fully compliant to the IETF specification. To fill this gap, I am going to implement a novel MIPv6 simulation model which can be integrated into ns-3. The project idea aims at testing the code in different network scenarios, containing different link layer technologies such as Wi-Fi, WiMAX and LTE. The current implementation of LTE does not have support of IPv6 in ns-3. So, the idea of testing MIPv6 code into LTE would follow IPv6 support implementation in LTE first and then the MIPv6 support. So, ns-3 researchers can benefit themselves with the addition of this feature. Thus, implementation of MIPv6 in ns-3 as the base mobility management solution and providing LTE support within it could help the current network researchers working on ns-3.</p>

Okular, the cross-platform universal document viewer developed by KDE supports PDFs with forms. These forms often use Javascript to make forms more convenient for its users. However, as of today, the support for Javascript within Okular is lacking. A lot of pre-defined javascript form functions are either partially implemented or not implemented. This often breaks form functionality for the users. Also at present, many essential objects and functionalities outlined in the PDF Javascript API Reference specification are either only partially implemented or completely absent. This leads to a lot of bug reports regarding broken form support for Okular. This project aims to resolve this, by improving the Javascript support for PDF forms. For this, the pre-defined Acrobat methods shall be implemented and newer JS objects shall either be added or improved upon during the course of this project. Deliverables : 1. Pre-Defined Acrobat method implementations. i. AFNumber_Keystroke ii. AFSpecial_Keystroke iii. AFMerge_Change iv. AFTime_Keystroke v. AFDate_Format vi. AFDate_Keystroke vii. AFPercent_Format viii. AFPercent_Keystroke 2. Expand the implementation for the event object. 3. Expand the implementation for the field object. 4. Implement the global javascript object. 5. Implement the submit and reset actions.

<p>This project aims to bring post-training integer quantization to OpenCV’s DNN module and perform inference using 8-bit integer inputs and fixed-point arithmetic. Using a simple quantize() function, any Net object can be quantized, resulting in roughly 1.5x faster inference, 4x reduction in memory consumption and accuracies close to floating-point inference.</p>

<p>SymbiFlow does not currently support partial reconfiguration regions. Partial reconfiguration regions are crucial to ongoing FPGA research including reducing verilog to bitstream compilation times through separate compilation. Two major hurdles to the support of partial reconfiguration regions are SymbiFlow and VPR support for restricted placement/routing and the ability to generate and upload a partial bitstream. I will be taking on the first of these two hurdles.</p>

<p>The Kubernetes dashboard previously supported Third Party Resources (TPR), but these were replaced in Kubernetes by Custom Resource Definitions (CRD). As a result, the original TPR support was removed in Dashboard, but CRD support has not been added yet. This proposal aims at providing a generic support for Custom Resource Definitions to the dashboard, similar to the previous TPR support.</p>

<p>The Common Workflow Language (CWL) aims to standardize workflow languages to execute command line tools on the cloud and on local servers. Apache Taverna supports specific formats for import and export of workflows, but currently, CWL workflows are not fully supported. During the GSOC, we intend to provide structural import and export of CWL workflows. In addition, we intend to provide a way to support some of CWL command line tools inside Taverna and to export Taverna activities.</p>

<p>Drakvuf-sandbox is a malware analysis tool that is used to analyze the behavior of the malware in a contained environment. It uses the Drakvuf engine as its backend for interacting with the malware and abstracts a lot of stuff out making analyzing malware easy and user-friendly.</p> <p>Currently, Drakvuf-Sandbox only supports Windows-based Malware Analysis. The upstream Drakvuf also lacks the proper Linux support. Therefore, this proposal focuses on implementing the support for Linux-based malware analysis.</p>

<p>LibRoadrunner is a high-performance SBML based simulator for Systems and Synthetic Biology that uses LLVM to generate very efficient runtime code. SBML Arrays package is under development and provides support for expressing an array of components of a computational biological model. Arrays make it easier to represent large complex regular systems in a standard way. Currently, iBioSim is the only tool that supports SBML arrays. The goal of the project is to implement “SBML Array” support in libRoadRunner and ensure compatibility between libRoadRunner and iBioSim so that models can be exchanged correctly.</p>

The project focuses on writing a Python library for GRASS GIS that utilizes EODAG API. Currently, GRASS uses different APIs for each satellite data provider. As a result, a lot of maintenance has to be done to keep up with the changes happening to all the different APIs. Thus, adding EODAG support to GRASS GIS will provide a lot of flexibility in the future development of GRASS, by decreasing the amount of maintenance needed to keep up with all the different APIs updated, and instead maintain the support for EODAG library. Deliverables for this proposal are supporting Sentinel, Landsat, and MODIS datasets with EODAG, and reimplementation of both i.sentinel.download and i.landsat.download modules with the new GRASS EODAG library.

MetaCall currently supports Linux (GCC), Windows (MSVC), and macOS but many platforms where developers want polyglot runtime support are missing.The core layers have partial support for some platforms but no CI, no environment script integration, and no testing to verify they actually work.Follow TDD approach: set up CI first , then fix environment script + build system, then fix test failures.I will go platform by platform for implementation.All platform logic lives in metacall-environment.sh there will be no hardcoded logic in CI YAML. 8 platform targets with CI pipelines with working environment script section, verified CMake detection and build, automated testing.Fixes to metacall/plthook fork for platforms that need it.