Find open source projects actively accepting contributors. Search repositories, filter by program milestones, difficulty tags, or tech stack.
Use our Orbit AI Matcher to find out! Get instant matching scores based on your developer skills, preferred frameworks, and contribution experience.
Convert your selected open-source project into a winning GSoC, LFX, or Outreachy application using Proposal Studio.
The AI-Powered Media Caption Generator is a smart Drupal module designed to automatically generate engaging and customizable captions for uploaded media using artificial intelligence. It leverages Google Vision API for image understanding and OpenAI along with Hugging Face Transformers for natural language captioning. Users can tailor the captions by choosing the tone, length (0–2000 characters), keywords, and emoji usage. A Python-based backend (using FastAPI or Flask) powers the AI caption generation and offers API support for optional integration with Drupal. The user interface will be built directly within Drupal to ensure a seamless media management and captioning experience.
Problem: Symbolic regression is crucial for discovering underlying physical laws from data, but traditional methods are often computationally expensive or lack generalization. Current approaches struggle with efficiency and adaptability across diverse scientific datasets. Solution: This project aims to develop a foundation model for symbolic regression by integrating transformer-based architectures (e.g., SymbolicGPT) with evolutionary algorithms and learned concept libraries (LASR). The model will leverage large-scale pretraining and fine-tuning to generalize across mathematical and physical datasets, combining the strengths of language models and evolutionary search for interpretable, efficient equation discovery. Deliverables: A transformer-based symbolic regression model with mathematical priors and curriculum learning. Integration of a learned concept library for reusable abstractions. Benchmarks against SOTA models. Optimized, modular code for seamless LLM integration and open-source deployment.
This proposal aims to enable the storage and retrieval of NDCube objects using the ASDF file format. Currently, ndcube lacks native support for saving NDCube objects to files and loading them back, limiting its usability. To address this, the project will implement converters and schema definitions for NDCube objects and related classes ensuring compatibility with ASDF. By leveraging ASDF's flexible structure and support for complex metadata, this project will enable seamless storage and retrieval of NDCube objects with their associated metadata, enhancing reproducibility and data management. Essentially, this project will enable any NDCube object to be saved to an ASDF file and loaded back in, making ndcube more user-friendly and improving how data is handled.
My project aims to enhance Chromium's AudioWorklet with robust test suites and developer resources for reliable real-time audio processing. I plan to achieve this by creating an automated test suite using Mocha and Chai, a live test suite with a publicly accessible UI, and a comprehensive starter guide for beginners. These resources will ensure the stability and performance of AudioWorklet, benefiting developers working with real-time audio processing on the web platform.
The project aims to finalize core functionality regarding initialization of problems and visualization of results using Paraview within the SU2 GUI built using Trame. Additionally, it will simplify installation via pip for running SU2 GUI. It aims to assist inexperienced users in setting up CFD problems efficiently in SU2
GSoC Project "Instrumentation of Clang/LLVM for Compile Time" aims to solve a simple but interesting question: Why is the compilation taking so long? We will improve the existing timing infrastructure to make the timing tools more efficient. Specifically, we improve the timing trace model to effectively time nearly all compilation stages. After that we can make statistical analysis based on the trace data, such as hotspots identification and so on. Finally, we implement the timing information collecting mechanism based on the library use case. This helps LLVM based languages or tools use related functions to easily troubleshoot its performance bottlenecks.
<p>Currently parameter configuration of DolphinScheduler is mainly based on configuration files. However there’s no way to override parameters when the same parameter is passed as JVM parameter at server startup. This project is about introducing parameter injection from java JVM arguments to server runtime, so that one can override a particular parameter which is externalised to a configuration file.</p>
<p>I want to improve the Linux kernel environment model for more accurate work with memory.</p>
<p>Selecting a good product for consumption is hard due to bad products in the market. So an application that can trace a product and its raw materials are the best options to have when shopping. S-ChainX is such an application. But it should integrate new features to make consumer's life easier. So Quick product search, Customer feedback, Court order links, Ingredients, and their effect on the human body are key facts that consumers need to aware of when buying a product. So in this project, I will try to add these enhancements to S-ChainX application. Also, I'm hoping to add unit tests and integrate the sonarQube to improve and maintain the quality of the code.</p>
<p>Field Officer Application is an application developed for the Bank staff field officer to keep track of Clients, Centers, Groups, Loan Account, Savings Account, etc. Currently, Mifos Android Client is in its Version 5. The app is still under development and this project aims at the release of Version 6 of this app.</p> <p>My goals will mainly include: adding T-OTP based 2-factor authentication using Google Authenticator app, deeper integration of Notifications framework, integration of SMS communications (in-app push notifications), UI improvements and redesigning and testing & documentation.</p>
<p>Amahi is a personal/Home server media software with the support of mobile apps and other devices that allow us to stream and access all types of media files present in our own servers (also called as HDAs). Currently, the android app works great, but there’s a lot of scope to improve it. This project mainly focuses on improving the following areas of the app:</p> <ol> <li>Enhancing the UI / UX improvements of the app.</li> <li>Adding support for secondary user login for both local and remote connection.</li> <li>Support to share folders and shares with other servers (Friending feature)</li> <li>Enhancing the media streaming of the app</li> <li>Enhancing the user experience of the app while accessing the files.</li> </ol>
<p>Bazel on Linux already supports various ways to run actions in a sandbox (linux-sandbox, processwrapper-sandbox). We want to support sandboxing on Windows by reusing the DetoursService from BuildXL (<a href="https://github.com/Microsoft/BuildXL" target="_blank">Microsoft Build Accelerator</a>).</p> <p>Related issue: <a href="https://github.com/bazelbuild/bazel/issues/5136" target="_blank">https://github.com/bazelbuild/bazel/issues/5136</a></p>
<p>My proposal is based on getting full support for JavaScript within the RoboComp framework. For this, the current state of generation of written components in the JavaScript language must be improved. Last year during the 2017 GSoC the functionality of the RoboComp component generator was extended to provide component supports based on NodeJS. A first approximation was achieved but still lacks robustness, reliability and should be extended to include more features. In addition, recently Zeroc ICE has added support for JavaScript interfaces, therefore the component generator must be extended to support this feature, both as a server and as a client. The component model of RoboComp must be able to generate components in this language with the same functionality as its counterparts in C ++ and Python, thus achieving complete support for JavaScript within RoboComp.</p>
<p>The project is about refactoring the current dual brain activity visualiser's 2D plots and adding new visual components to its 3D part to display more information about the signals. We will rewrite the 2D part using a modern data visualisation framework. In the 3D part, apart from the instant activity of the brain, we will add a dynamic sphere at the location of the signal. The proposed new viewer will also have the interactivity of zooming, play and pause for the users.</p>
<p>This project will add a package manager to Blender’s current addon system with the goal of providing functionality similar to that found in <a href="https://atom.io" target="_blank">Atom</a> (the text editor).</p>
<p>This project is about improving LLVM backend for chapel compiler. Improvements consist of improving loop vectorization by adding proper metadata and changing loop compilation, adding type-based alias analysis metadata, adding notations when there is load from const memory, experimenting with aliasing algorithms, adding polly optimizer and extending support for wide pointers. Main goal of the project is to improve loop vectorization.</p>
<p>Beamforming is among the most widely used source localization techniques for MEG and EEG data in neuroscience. This project aims at porting an updated version of the five-dimensional beamformer to analyze time-frequency data in source space (Dalal et al., 2008; implemented in MNE-Python as “tf_lcmv”). The update adopts the Hilbert transform to obtain a continuous amplitude and phase time series per frequency band, in contrast to the previous method that utilizes discrete time windows and does not retain phase information. It also adds a weight normalization option to the LCMV beamformer (neural activity index and unit noise gain), which will be available for other LCMV versions as well. Furthermore, statistical testing of the Hilbert beamformer output using the Wilcoxon rank sum test will be implemented to complete the new Hilbert beamformer pipeline. Due to the size of the MEG data and intensiveness of the calculations, the code will be developed with memory efficiency and parallelization in mind. Interactive visualization of source-localized time frequency representations will complete the Hilbert beamformer routine.</p>
<p>Development of a deep learning framework to discover personalized genomic medicine for lung and breast cancer</p> <p>I’d like to solve the above problem with supervised, unsupervised, deep, and wide models. Models I’d like to pay special attention to are recurrent neural networks and hidden markov models. I will read relevant papers for reference, and will implement based on them. I'd also like to apply some linear algebra techniques for better feature extraction. Some of the papers that I’d like to refer to are like the following:</p> <p><a href="https://www.ncbi.nlm.nih.gov/pubmed/17069508" target="_blank">https://www.ncbi.nlm.nih.gov/pubmed/17069508</a> <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4965871/" target="_blank">https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4965871/</a></p>
<p>Liberty Eiffel language available on windows environment.</p>
<p>There is functionality for working with dynamical systems over projective space in Sage. However, one of the areas lacking in functionality is the moduli space. We say two self-maps of projective space are equivalent if there is an element of the Projective Linear Group that conjugates one to the other. The following two algorithms should be implemented, in order to create more functionality in moduli space. The first algorithm would be that given two endomorphisms of projective space determine if they are conjugate. In other words, determine if they are in the same class in the moduli space. If they are, also return the PGL element that conjugates one to the other.The second algorithm would be that given an endomorphism of projective space to compute a reduced form. This conjugation would that makes the coefficients small. There is already an algorithm implemented to return the minimal model in terms of resultant, but the coefficients can be non-optimal. The simplest approach would be to reduce the binary form describing the fixed points or if the binary form describing the fixed points is not sufficiently nice then reduce the binary form for points of period n for some small n.</p>
Open Rocket.Chat communities face a growing problem where new accounts join, flood multiple public channels with spam, and disappear before any moderator can respond. Current moderation is entirely reactive and manual, which is simply too slow for coordinated spam campaigns. This project builds a proactive, human-validated anti-spam system that monitors new users during their first 6 to 10 weeks and responds through a graduated restriction system called Moderation Levels, where every action is proportional, reversible, and requires admin approval before anything permanent happens. The system is built as a Rocket.Chat App on the Apps Engine with a seven gate detection pipeline running entirely in memory for speed. Gates catch exact duplicates via hash matching, polymorphic spam via cosine similarity, cross channel floods, rate flooding, rapid room spread, and phishing link campaigns. A four tier Moderation Level system escalates naturally with repeat behavior, from a friendly DM warning through posting cool downs to a permanent block that only lifts when an admin vouches for the user. The admin control layer includes a live dashboard, a full slash command suite, and AI powered on demand summaries so admins have everything they need to make confident moderation decisions without opening a single log file. Deliverables include a fully functional installable Rocket.Chat App covering the complete detection pipeline, a configurable Moderation Level system, a behavioral scoring engine, enriched admin notifications with message history, AI powered user analysis commands, a paginated admin dashboard built for scale, daily scheduled reporting, and complete documentation covering architecture, configuration, and known limitations.
VLC uses lua scripts for tasks like playlist parsing, service discovery, and art fetching. Currently, testing these scripts requires manually running the full VLC application. This is inefficient and fails to catch bugs when external websites change their markup, leaving scripts broken. This project builds a standalone CLI test harness (vlc-lua-mock) that replicates the VLC Lua engine's behavior without requiring a VLC build. It uses mock backends and stubs for VLC Lua bindings, supports both behavior and state assertions, and includes a recording mode for fixtures.
Parallelize Navier-Stokes hypersonic shockwave solver kernels using OpenMP 5.0 target offloading on European supercomputer clusters.