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.
<p>Since the release of CC Search Browser extension, it has garnered a good number of users. But, the extension still does not have sufficient features and capability to be used as a daily driver in the workflow of the users and therefore most of them use it as a temporary plugin to search for CC-licensed images. My main aim would be to transform the extension to become competent enough to contribute to the user's workflow and be an important part of their toolkit. I will do that by adding new features in addition to other enhancements and improvements to the extension.</p>
<p>Syslog-ng is a system logging application available for a wide range of operating systems. It is capable of sending, filtering, parsing, rewriting and forwarding logs to various destinations. It runs as a daemon process on systems and configured via files.</p> <p>The projects goal is to extend its functionality by making it able to run offline as a command line tool reading incoming logs from stdin, transforming the input the usual way and forwarding it to stdout.</p>
This proposal provides a detailed overview of the development of an LLM-based conversational assistant with domain-specific knowledge about the OpenROAD project. The primary focus is on data engineering and testing to enhance the model’s capabilities. The assistant is designed to improve user interaction by providing accurate responses to queries regarding OpenROAD’s tools and features. We will be leveraging Fine-Tuning and Retrieval-Augmented Generation (RAG) architecture and we will address key LLM challenges.
The goal is to implement a functional GNSS receiver that is capable of introducing measurements from another sensor into the GNSS receiver chain. This will advance research on sensor fusion, enabling reliable high-accuracy navigation solutions in challenging environments such as the urban cannon. The end product will consist of a SignalSource implementation, responsible for talking with the sensor at a hardware level and synchronizing its measurements with GNSS samples; and the acquisition, tracking and PVT algorithms implementing the sensor fusion model.
This project aims to improve the previous iteration of the project where the hypothesis of it was the existence of a relation between the body gestures and the time expressions we speak and using neural networks we try to accept or reject this hypothesis. So, I proposed a new multimodal graph-based neural network architecture to classify the body keypose points by constructing a spatio-temporal graph and audio features are fused with it to converge the model to test this hypothesis. Also, suggested transfer learning to overcome the lack of data.
The project aims to develop a comprehensive prompt editor within Rocket.Chat, facilitating seamless interaction between users and open source language models (LLMs) such as Mistral, Llama 2, and Phi. Key objectives include enabling free-flowing conversations, robust conversation history management, and seamless prompt sharing across external applications. The initial focus will be on implementing this functionality within the Rocket.Chat Web App and Electron platform.
Discovering new phenomena, like the Higgs boson, involves the identification of rare signals that could shed light into unanswered questions about our universe. On the other hand, at the Large Hadron Collider (LHC), large amounts of data are generated from particle collisions. This project aims to develop a Quantum Lie-Equivariant Graph Neural Network (QLie-EGNN), a quantum model that not only is data efficient, but also has symmetry-preserving properties from any arbitrary Lie algebra that can be learned directly from the data.
<p>This project aims to improve the Reference Image Tool in Krita to include some additional features that will be helpful for the users .With this project , users would be able to directly interact with the reference images and modify them accordingly with Krita . It includes</p> <p>Pin Reference Image - a new transformation to prevent canvas modifications to them .</p> <p>Integrate crop functionality for them .</p> <p>Update & sync the reference image when image is modified on the hard drive .</p>
TensorFlow Decision Forests (TF-DF) is a collection of state-of-the-art algorithms for the training, serving, and interpretation of Decision Forest models. Unfortunately, the lack of detailed examples in the community is preventing many people from fully realizing its potential. In this project, I will introduce a series of examples that will demonstrate the use of TensorFlow Decision Forests on various datasets. I’ll dive into beginner as well as intermediate examples that will help improve the performance of these tools on structured datasets.
<p>Casbin is an open-source Authorization Library which supports access control models like ACL, RBAC, ABAC for Golang, Python, C/C++ and for many other languages. There are many casbin plugins which support different tools. Kubernetes being one of the biggest project in Go didn't have this support. So, this project is about developing a Casbin plugin for Kubernetes, which would authorize the users based upon the roles/policies defined in the Casbin policy file to access certain pods or clusters.</p>
<p>In our work, we aim to research, implement and integrate a method that is capable of real-time manipulation of faces on mobile devices like changing the style of the hair, color of the lipstick or eyes, increasing the length of the hair, adding or removing facial hair, adding or removing smile and similar. We aim to be able to perform these kind of transformations while also preserving the identity of the input face and only changing local part of the input image of face while keeping other parts untouched.</p>
<p>The PowerUp-Android is an Android application for empowering preadolescent girls between the ages of 12 to 14. The game is based on Social & Emotional Learning model to help girls to learn conflict resolution, negotiation & other social skills needed to build their confidence against tough situations. The user creates a character and tours through different scenarios each dealing with the different subject of importance. These scenarios contain mini-games which are launched between the scenarios to keep user the engaged.</p>
Proposing an immersive augmented reality application that transforms marine biology education through interactive experiences where users can embody marine creatures, particularly octopuses, to understand marine ecosystems. Leveraging my experience in Unity AR development, Android native programming, and back-end integration, I aim to create a comprehensive learning tool that combines gesture recognition, voice commands, and realistic 3D environments to make marine biology accessible and engaging for students of all ages.
MathOptAI.jl is a Julia package for embedding machine learning models into optimization problems built using JuMP. These embeddings work as optimization surrogates and replace the terms in the original formulations with approximations represented by the added variables and constraints. Input convex neural networks have some useful properties such as convexity that enable approximations with similar quality with much simpler embedded formulations. This GSOC project will extend MathOptAI to add support for input convex neural networks.
Recently, research has been made into the process in which we could tune an existing generalized LLM (such as ChatGPT 3.5, 4) for the purposes of generating HDL (especially Verilog/Chisel) code from English. However, we still need a method of evaluating these models so that we can assure that they have full coverage over the language. To this effect, we create multiple large benchmarks in both Verilog and Chisel along with a respective set of prompts. In addition, we produce a program to test the syntax, functionality, and quality of the generated code against our benchmarks.
<p>Toolkit for Multivariate Analysis (TMVA) is a multi-purpose machine learning toolkit integrated into the ROOT scientific software framework, used in many particle physics data analysis and applications. A fast inference system that takes takes ONNX model as input and produces compilation-ready standalone C++ scripts as output is under development in TMVA. This project will focus on the development of ONNX RNN, LSTM and GRU operators in the code generation format.</p>
This project aims to address the challenge of text recognition from historical Spanish printed sources dating back to the seventeenth century, a domain where existing Optical Character Recognition (OCR) tools often fail due to the complexity and variability of the texts. Leveraging hybrid end-to- end models based on a combination of multiple architectures, such as CNN-RNN, our research seeks to develop advanced machine learning techniques capable of accurately transcribing non-standard printed text.
<p>This came from a project I pitched last year with Antranig Basman and Colin Clark. Our goal is to make public data sources accessible for analysis and visualization by creating an end-user application that allows users to choose from data sources, do transformations on that data, and ultimately create shareable visualizations of the data. The goal is to make a rapid visualization framework that is accessible and has many available forms for the final representation including sonification to open representations of data up to the visually impaired.</p>
<p>The performance of the code generated for the GPU heavily depends on runtime parameters such as the number of thread blocks and threads per thread block. For this reason, a proper choice of such parameters based on both the IR information and runtime information can give a tangible improvement of the performance of the generated code. In this project, we will explore the ML-based approach to this problem which uses machine learning models for making a decision on the runtime parameter values given the IR and runtime features.</p>
<p>The prior goal of this project is to implement a new fairness package that is well integrated into the mlr3 ecosystem. This new package could be used by many other mlr3 subsystems like mlr3pipelines, mlr3tuning and mlr3proba. Then further help the researchers and industrial users to detect the fairness of datasets or algorithmic bias more easily, which could benefit the accuracy and impartiality of their models built by mlr3. The project may also involve some debiasing algorithms in recently published papers due to the connection to mlr3's tuning facilities.</p>
<p>Deepbots is a wrapper framework for the open-source robotics simulator Webots. It acts as a middleware between the simulator and Reinforcement Learning use cases. For eg. it can be used to implement and test different Reinforcement Learning algorithms by using the different robot and world models as an environment. The API is very similar to the standardly used OpenAI gym environments. The goal of this project is to extend deepbots to integrate and facilitate implementation of different evolutionary algorithms and test them in Webots using deepbots.</p>
The proposed solution introduces comprehensive tracing capabilities to the LLM4S toolkit by integrating OpenTelemetry (otel4s) for capturing detailed spans and metrics during LLM operations such as prompt construction, model invocation, and response parsing. These telemetry signals are exported in Prometheus and Jaeger-compatible formats and processed through the OpenTelemetry Collector. The data is then visualized on a unified Grafana dashboard, allowing developers to monitor performance, debug errors, and analyze execution flows across different LLM providers in real-time.
The current proposal deals with the task of document similarity utilizing dense text representations. Furthermore, it leverages topic modelling to investigate potential non-informative sets of words which could be removed in the preprocessing phase. The latter step will improve training time and hopefully will increase performance on the document similarity task. The task occurred from the need of matching Github issues with the relevant Jenkins logs in order to correlate ci pipeline errors with prior identified issues.
Planets form in complex, dynamic environments. The physics behind the process is not well-understood, and the limited high-quality observational data is difficult to interpret. Using high-fidelity hydrodynamical simulations of protoplanetary disk evolution under a variety of physical and observational conditions, machine learning models will be trained that can identify and quantify the physical processes that influence the protoplanetary disk. This will speed up and improve the analysis of observations, furthering our understanding of the environments and processes that are responsible for planets.