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.
GeNN is a C++ library that generates code for efficiently simulating Spiking Neural Networks (SNNs) using GPUs. In order to compile this generated code, it requires a compiler and development versions of backend requirements such as CUDA. Because of this, GeNN must currently be installed from source which puts off some users. So for this project, I aim to develop a conda (forge) package for GeNN which will handle the installation of the required dependencies and ease installation for end users. Deliverables: -Conda Forge Recipe -User Documentation
I plan to develop 10 Math games for Sugar within the GSoC’25 timeline. The development of these activities will include basic functionalities of the game in addition to other features which are not mentioned particularly like Sound and themes and sugar features like journals (to keep the changes saved for game resume features), user color schemes, toolbar, etc
Modern RTL simulation is computationally expensive and often limited by single-node processing capabilities. This proposal aims to improve Metro-MPI++, a tool for distributed RTL simulation using Verilator and MPI, to better handle large-scale designs like OpenPiton. The project focuses on automating tile and interconnect detection in OpenPiton, optimizing MPI-based simulation interfaces, integrating hybrid parallelism with OpenMP, and laying the groundwork for FPGA backend support. The deliverables include a scalable partitioning framework, enhancements to the simulation runtime, example simulations demonstrating speedup, and contributions to Verilator’s partitioning features. These improvements will make RTL simulation more scalable, accessible, and efficient for open-source hardware projects.
BugSim (Bug Simulator and Learning) is an advanced bug simulation platform integrated into OWASP BLT. It helps security researchers, ethical hackers, and developers understand, simulate, and resolve vulnerabilities in real-world environments. The project includes features like: Hands-on bug simulation and learning with various vulnerability categories (SQL Injection, XSS, CSRF, etc.). Weekly vulnerability challenges and structured bug reporting. Integration with OWASP ZAP for automated security testing. Live hackathons and competitive bug hunting with leaderboards. Gamification with points, badges, and rewards. This project will enhance OWASP BLT by creating an engaging and interactive cybersecurity learning platform. The deliverables include a fully functional BugSim module, complete with vulnerability simulations, a scoring system, and real-time security reporting.
Perspective addresses the problem of biased and one-sided narratives in online articles by providing readers with a fact-based, well-structured alternative perspective. The solution involves a pipeline that scrapes article content, processes and cleans the text, extracts key points, and generates a logical counter-perspective using LangGraph and LangChain. The system will include a frontend for user interaction, an AI-driven backend for perspective generation, and a vector database for storing results. Deliverables: - A fully functional web application where users can input article URLs. - AI-powered alternative perspective generation based on extracted key points. - Fact-checking mechanisms using reliable sources. - Vector database storage for efficient retrieval and chat-based exploration.
refer notion site for better view : https://smart-rhubarb-c5c.notion.site/GSoC-2025-Proposal-Project-12-Empowering-Kubeflow-Documentation-with-LLMs-1c209ac9a7bd80e6941ee51ab6601c7c This project aims to build an AI-powered assistant for the Kubeflow ecosystem using Retrieval-Augmented Generation (RAG) and a long-context LLM. The assistant will index Kubeflow Github repository including, GitHub issues, and community discussions using a structured pipeline involving Feast (feature store) and Milvus (vector DB). It will support real-time queries through a FastAPI WebSocket interface and serve responses via KServe. Key deliverables include: End-to-end RAG pipeline integrated with Kubeflow Pipelines & Feast GitHub-triggered document ingestion and vector index updates LLM deployment and retrieval API with RBAC controls Evaluation loop using “LLM-as-a-Judge” Monitoring setup with Prometheus + Grafana Frontend integration with Hugo-based docs site This project will improve Kubeflow’s developer experience by enabling contextual, cited answers directly from its ecosystem.
This document outlines a proposal to support Volcano for gang-scheduling in Kubeflow Trainer, so as to provide users with more AI-specific scheduling capacities like priority scheduling and queue resource management. This will allow for better control over resources, with features like queue management and job prioritization. Thanks to the Kubeflow Trainer Pipeline Framework, we can seamlessly integrate Volcano into Kubeflow Trainer as a runtime plugin. The deliverables include: 1. The integration of the Volcano plugin into the Trainer, allowing for Gang-Scheduling and resource management. 2. Documentation updates, including usage examples and design details. 3. Testing the plugin's functionality and performance under various environments.
Synopsis: This project aims to enhance the brainglobe-registration tool by automating the selection of atlas regions for aligning 2D or 3D brain images to a standard anatomical reference via brainglobe-atlasapi. Currently, this process is manual, time-consuming, and subjective, posing a barrier to reproducibility and efficiency. By introducing a new preprocessing step that automatically identifies the optimal atlas region – using methods such as adaptive grid search, machine learning, or Bayesian optimisation – this project will streamline the workflow and improve accuracy. Deliverables include at least two implemented approaches, benchmarking, comprehensive testing, full documentation, and a blog post to showcase the new functionality. This automation directly benefits my research in paediatric radiotherapy, where mapping histology images to an atlas is essential for analysing cellular behaviour across brain regions. By making this process faster and more robust, it not only enhances my own work but also enables broader adoption of BrainGlobe tools by researchers in neuroscience and healthcare, even those without extensive anatomical expertise. Minimal Set of Deliverables: • Implement a new preprocessing step to automatically select the atlas region. • Compare at least two region-selection methods (e.g., adaptive grid search vs ML/Bayesian optimisation). • Add tests to validate the new functionality. • Document the new pipeline and add usage instructions. • Write a blog post showcasing the project and results.
OWTF's built-in MiTM proxy was developed nearly a decade ago and is currently outdated, limited in functionality, and hard to maintain. This project aims to completely modernize the MiTM proxy component to make it as feature-rich, stable, and flexible as modern alternatives like mitmproxy. The upgraded proxy will support TLS interception, live modification of HTTP(S) requests and responses, robust transaction logging, and seamless integration with the OWTF framework and frontend. Deliverables include a fully functional, test-covered modern proxy component with improved logging, dynamic tampering support, and compatibility with OWTF's plugin and reporting infrastructure.
The project involves integrating pixel size support into ilastik, an image-processing and classification tool. Pixel size is a metadata item enabling conversion of image pixels into real-world measurements. Whilst pixel size is not currently processed in ilastik image sets, the program is used extensively in scientific applications, and introducing real-world measurements would enable development of several key features that would allow more in-depth quantitative analysis of image data, such as object displacement and velocity in video tracking, and real-world area calculation of segmentations/ volume calculation of carvings. Once the metadata is extracted at the image input end, it is possible to feed it directly into the data pipeline, so it can be picked back up within any relevant workflow. This facilitates feature development, as it then becomes possible to work directly with the metadata values within the workflow. Once the backends of these features have been developed, they will be implemented on the front-end via Qt. The key deliverables of this project are: Implement pixel size metadata into workflows, such that the metadata remains present in project files, output images, and exported datasets; Enable calculating areas of image segmentations; Enable calculating volumes of image carvings; Enable recording of object displacement in video tracking; Enable measurement of object velocity in video tracking; Produce detailed and well-written documentation, and ensure relevant test scripts for each feature are made available. An initial pull request for this proposal can be found at https://github.com/ilastik/ilastik/pull/2995
This project suggests significant upgrades to Salsa CI (Debian's Continuous Integration system) used by more than 27,000 source packages by adding an effective statistics dashboard, resolving issues, making the pipelines faster, and doing smart job skipping for non-functional changes. Altogether, these enhancements are hoped to streamline the CI process, save resources, and empower Debian maintainers with improved insight, quicker feedback loops, and a smoother development experience.
This project aims to finalize the system for verifying the integrity of game files within ScummVM. It introduces a server-client architecture where hash-based checksums are computed for game files and compared against a curated database. The system provides moderators tools to validate, merge, and maintain the database. A major focus is on automating the pipeline, enhancing moderation tools, and addressing edge cases.
The aim is to refine the capabilities of asynchronous I/O within NetBSD. A program that utilizes asynchronous I/O is able to invoke multiple file operations and continue executing without blocking on their explicit completion. The guiding principle is to keep submission cheap, coalesce work sensibly, and only spawn worker threads when the kernel would otherwise block. Today much of the path within the kernel still behaves synchronously; even so there are performance gains when working with different files, and the work lays the groundwork for future optimizations in concurrency on the same file. We are investigating practical paths forward by drawing on designs that decouple submission from waiting so the system may stall less, with a clear route towards maximal parallelism as lower layers mature (e.g. the buffer cache).
CAR T-cell therapy is a form of cancer immunotherapy that engineers a patient’s T cells to recognize and eliminate malignant cells. Although highly effective in leukemias and other hematological cancers, this therapy faces significant challenges in solid tumors due to the complex and heterogeneous tumor microenvironment. CARTopiaX is an advanced agent-based model developed to address this challenge, using the mathematical framework proposed in the Nature paper “In silico study of heterogeneous tumour-derived organoid response to CAR T-cell therapy,” successfully replicating its core results. Built on BioDynaMo, a high-performance, open-source platform for large-scale and modular biological modeling, CARTopiaX enables detailed exploration of complex biological interactions, hypothesis testing, and data-driven discovery within solid tumor microenvironments. The project achieved major milestones, including simulations that run more than twice as fast as previous model, allowing rapid scenario exploration and robust hypothesis validation; high-quality, well-structured, and maintainable C++ code developed following modern software engineering principles; and a scalable, modular, and extensible architecture that fosters collaboration, customization, and the continuous evolution of an open-source ecosystem. Altogether, this work represents a meaningful advancement in computational biology, providing researchers with a powerful tool to investigate CAR T-cell dynamics in solid tumor and accelerating scientific discovery while reducing the time and cost associated with experimental wet-lab research.
The 2D Arrangements package in CGAL provides an efficient and extensible framework for representing and manipulating arrangements induced by geometric objects. The existing visualization utilities are useful but exhibit several limitations: inefficient rendering procedures of faces with holes, lack of support for unbounded curves and faces, and inability to correctly render faces on curved surfaces. This project implements a general renderer for arrangements that integrates with the Aos2 framework and addresses those limitations. The new renderer provides on-demand approximation of x-monotone curves; uses a customized face triangulator to efficiently produce meshes; and leverages tiling and caching to make interactive navigation responsive.
This project aims to develop a cross-platform application for Mankala using Qt and QML, extending the capabilities of the existing MankalaEngine maintained by the KDE Community. The application will feature a responsive, touch-friendly interface and support both local gameplay modes and real-time multiplayer using the Extensible Messaging and Presence Protocol. By combining a modern UI with networked communication, the project seeks to bring traditional Mankala gameplay to mobile devices while promoting open-source tools and accessibility.
Problem: While Large Language Models (LLMs) like ChatGPT can generate R code, they often lack precision when answering questions about third-party R packages or contributing to the R language itself. This can limit their usefulness for R users, particularly for new contributors seeking to understand the R core or engage in package development. Furthermore, many of these AI tools are proprietary or resource-intensive, excluding users with limited hardware or funding. Solution: ChatR proposes a specialized R chatbot that is lightweight, locally runnable (even without a GPU), and customized to provide reliable support on R packages, documentation, and contribution practices. The project will use Retrieval-Augmented Generation (RAG), prompt engineering, and a lightweight open-source LLM (via Ollama and llama3.2:3b-instruct) to deliver targeted responses. It will be delivered as an R package with both CLI and Shiny UI interfaces. Deliverables: An R package wrapping the chatbot and customizable prompts Local model integration using Ollama and ellmer RAG-based indexing of documentation (e.g., CRAN, Writing R Extensions) A Shiny widget interface for interactive usage in RStudio Documentation, test suite, and demo applications This project will significantly support learners, developers, and contributors by democratizing access to reliable AI-based R support and onboarding tools.
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.
Currently, ceph-devstack runs a testnode container with only one OSD per node using a loop block created on the host system as a storage device. To be closer to a production-like environment, it is important to make ceph-devstack run many OSD per testnode. To do that we need to : enable support for arbitrary number of OSDs and define size of storage devices. The deliverable will be a new version of ceph-devstack with listed features.
Traditionally, brain activity has been analyzed independently from peripheral physiological measurements like the electrocardiogram (ECG), but our understanding of brain-heart interactions (BHI) has significantly evolved since then: we now know that not only does the brain modulate the heart but surprisingly the heart exerts a powerful influence on the brain. For example, the timing of a learning task relative to the cycle of the heartbeat modulates its event-related potential (ERP), and heart-beats interact with activity in the insula, an area critically involved in internal monitoring. However, despite this increased interest in BHI in the neuroscientific community, there is a lack of well-documented tool, that incorporates much of the modern BHI identification techniques, treats sensitively ECG artifacts, and furthermore complements modern brain connectivity analysis. BHI studies point to a much greater complexity in neural signals that current connectivity packages fail to account for, such as recent work pointing to the modulating of brain oscillations by respiration. As such, this project has two main purposes: extract BHI and direct respiratory modulation of the brain and build a brain connectivity tool that accounts for these confounding measures, built in the Python MNE environment to complement traditional data analysis pipelines. Our analysis functions will furthermore be validated through tutorial-like experiments, written as Jupyter notebooks on open datasets, and we hope that the results I show will convince the scientific audience of the importance of confounding peripheral measures in neural data. The minimal set of deliverables include: individual analysis functions for peripheral physiological signals, connectivity measures between peripheral and neural signals, and brain connectivity measures, both regular, and conditioned on peripheral physiological signals to remove their effect, and experiments demonstrating these tools on open datasets.
Diffusion models have experienced rapid growth in usability, availability, and research. The classical algorithm consists of two main parts: a forward diffusion process and a learned denoising step. The forward diffusion process discretely applies noise to an image until it is unrecognizable. The denoising process reverses the noise, predicting the original image. Quantum techniques, such as parametric quantum circuits, could be applied to both parts to create a fully quantum model that could run on a quantum computer, speeding up training time. This model would be trained on multiple high energy physics datasets, such as the quark and gluon dataset for generating quark and gluon jets, alongside a classical diffusion model, and the results could be evaluated taking into account resources, time, and quality. Furthermore, a mathematically-driven approach would provide the theoretical foundations of quantum diffusion models in addition to experimental results.
This project extends capa, FLARE's open-source capability detection tool, with Android dynamic analysis using Frida instrumentation. The implementation automates the complete Frida workflow from template-driven script generation to feature extraction. This brings capa capability detection to the broader Android dynamic malware analysis community.
This project proposes integrating ACTINN, a neural network for automated cell type identification in single-cell RNA sequencing (scRNA-seq) data, into the DeepChem framework. Integrating ACTINN into DeepChem would mark the framework's inaugural inclusion of a single-cell deep learning model, significantly expanding its capabilities in the burgeoning field of single-cell analysis. By incorporating ACTINN, DeepChem can offer researchers a powerful tool for automated cell type identification, streamlining workflows in single-cell analysis and fostering deeper insights into cellular heterogeneity.