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.
FreeCAD comes with a Robot Workbench, which is capable of 6-DOF robot simulations, trajectory planning and a vendor specific code export (KUKA). After the removal of some copyrighted model assets and years without an active maintainer, the workbench is functional but difficult to use and customise. The robots can only be defined via hand-written VRML and CSV/DH files, the bundled kinematics library has lagged behind the latest upstream release, the axis count for robots is hard coded to six and the documentation for beginners is sparse. Thus this project proposes an incremental modernization of the existing “Mod/Robot” module: 1. Standardised robot imports : Unified Robot Description Format (URDF) is one of the most common standards for describing a robot. Having an easy way to import robots directly from URDF will replace the painful CSV/DH file imports. 2. Generic joint-trajectory export : Having the possibility to export the joint & trajectory data will separate the export process from copyrighted vendor-specific pipelines. The users can post-process the exported data to their specific platform 3. Kinematics library refresh : FreeCAD currently has a local fork of KDL library that lags behind the upstream. Cleaning up this dependency will eventually prepare the path for arbitrary-DOF robots & better robot animations. 4. Documentation : This will help the existing users as well as attract a new user base towards the Robotics Workbench.
Problem: Imbalanced datasets pose a significant challenge in machine learning, where standard metrics like accuracy can falsely evaluate a model that simply predicts the dominant class. Currently, the R community lacks a comprehensive and unified framework for handling imbalanced data and X-Risks problems natively integrated into the mlr3/mlr3torch ecosystem. Solution: This project, mlr3torchAUM, provides a comprehensive solution for imbalanced classification in R. It will bridge the current gap by integrating state-of-the-art loss functions, advanced evaluation metrics, and various data sampling techniques directly into the mlr3/mlr3torch pipeline. Deliverables: 1. State-of-the-Art Loss Functions: Implementation of algorithms including log-linear all-pairs squared hinge loss, Area Under Min (AUM) loss, and cost-sensitive logistic loss. 2. Sampling Operators: A robust suite of over-sampling techniques including SMOTE and ADASYN, alongside under-sampling techniques, all implemented safely under a `lazy_tensor` mechanism to avoid memory issues. 3. Advanced Evaluation Metrics: Integration of Precision-Recall curves, Macro/Micro AUM/AUC, and MCP/IMCP to fully support multiclass imbalanced scenarios. 4. LibAUC Integration: Porting cutting-edge losses and optimizers from the Python LibAUC library into R. 5. Engineering & Release: Comprehensive Test Driven Development, GitHub Actions for CI/CD, detailed vignettes, and a final submission to CRAN.
<p>The project aims to enable the Moodle Mobile App with fully automated End-to-End testing capabilities with SauceLabs and Travis CI using latest upgraded versions of Node, Protractor, Gulp, Appium, Selenium with latest Moodle Mobile App v3.3.0 [~15/5/2017].</p> <p>Integrating SauceLabs and TravisCI enables a completely automated build on Travis along with End-to-End testing on SauceLabs’ Browsers, Mobile/Tablet Android Emulators, Mobile/Tablet iOS Simulators [and Real Devices exclusively for premium users]. In case of any failure in Travis build, debugging can be easily done as SauceLabs provides the actual video and screenshot logs of all the tests for 30 days [only metadata & screenshots after that] and TravisCI restores the build log for the lifetime of a repository.</p> <h6>Project Goals</h6> <ul> <li>Upgrade Node, Protractor, Gulp, Appium, Selenium and other related dependencies</li> <li>Configure the App for Travis build and SauceLabs testing via Travis</li> <li>Fix all the failing E2E tests</li> <li>Add E2E tests for newly added features/modules in Moodle Mobile v3.1.x, v3.2.x, v3.3.0</li> <li>Modify the existing documentation of Moodle Mobile Development Environment Setup and End-to-End testing environment</li> </ul>
The ROS 2 ecosystem has made significant leaps in performance over the last few years, supported by benchmarking tools like buildfarm_perf_tests and ros2-benchmark-container. These tools take care of profiling middleware and client library performance under various configurations. As of now, the recurring monitoring path in continuous integration (CI) is mostly rclcpp-centric, providing good regression visibility for C++ implementations. However, a current limitation of the ecosystem is that it lacks periodic performance monitoring for other heavily used client libraries like rclpy (Python) and rclrs (Rust). This leaves these libraries with weaker regression visibility, making it difficult to measure the performance impact of new topologies involving remote host subscriptions or track the performance of ROS actions and services reliably. This project focuses on vastly extending the suite of tests that run regularly in buildfarm_perf_tests by porting C++ benchmark instrumentation for use with the other client library implementations. This project thus aims to add high-signal, recurring CI coverage across rclcpp, rclpy, and rclrs with minimal CI cost overhead. Furthermore, to enhance the contribution, there will also be a focus on updating the performance test result graphs in CI to use modern observability tools. The final deliverable includes an updated web dashboard (built with Grafana) that tracks key compute metrics such as latency, CPU usage, and memory over time, providing the community with a complete, cross-language performance diagnostic ecosystem.
The Nix build hook mechanism has been marked for removal since 2017, but it still fills an important gap in Nix's distributed build architecture. In particular, it is currently the only interface that can pass a build request together with its full set of input paths atomically. Store-based remote building does not provide this: it splits the process into separate calls, which makes it difficult for builder stores that schedule builds across a cluster, such as "nix-scheduler-hook", to know where inputs need to be copied before the build is assigned. This project proposes a cleaner, more extensible replacement by adding a new virtual "buildDerivation" method to the "Store" base class. This method would bundle the derivation path, derivation, and input paths into a single call, preserving the atomic behavior that cluster schedulers rely on. To ensure compatibility, the new interface would include a default implementation so that existing store backends continue to work unchanged. The project will validate this design through an end-to-end proof of concept by adapting "nix-scheduler-hook" to use the new API. Once the new mechanism is confirmed to support distributed scheduling correctly, the legacy build hook mechanism can be removed from the Nix daemon. Expected deliverables include the new virtual "Store" method and its default implementation, full removal of the build hook mechanism from the Nix daemon, a proof-of-concept "nix-scheduler-hook" adaptation, a passing distributed build test suite, and accompanying API documentation and migration guidance.
Open Food Facts (OFF) relies on contributions from a global community of developers and volunteers. However, setting up development environments, navigating complex CI/CD pipelines, and contributing to multiple repositories like off-server, robotoff, and smooth-app can be time-consuming and frustrating—especially for newcomers. This project aims to streamline the entire contributor journey by creating a unified, cloud-powered Developer Experience (DevX) platform for OFF.\n \n We will implement pre-configured Google IDX, Gitpod and Codespaces environments, enabling contributors to start coding within minutes. The CI/CD pipelines will be optimized to run only affected modules, significantly cutting build and test times. We'll also create a live, versioned documentation pipeline, integrate AI-powered tools for issue triage and PR review, and develop a metrics dashboard for maintainers to track workflow efficiency.\n \n Deliverables include: \n - Deployment staging environment with Docker swarm or Ansible.\n - Auto-launch Gitpod/ Codespace environments. \n - Optimized CI/CD workflows with modular test execution. \n - Automated documentation site using OpenAPI Spec + MDX + Fumadocs. \n - AI bot integration for issue triage. \n - A real-time DevX dashboard. \n \n This project will empower contributors to go from zero to green PRs faster, making OFF development more accessible, efficient, and scalable.
<p>Main goal: have the AD7292 driver in the Linux kernel tree.</p> <p>Secondary goal: develop a series of tutorial posts for newcomers.</p> <p>The following summarizes the deliverables I want to provide.</p> <ul> <li>AD7292 driver</li> </ul> <p>Develop the AD7292 driver incrementally adding small pieces of code every week and gathering feedback from the community during this process. This deliverable will be the sum of several minor contributions along the entire GSoC period.</p> <ul> <li>Post about AD7292 structure overview</li> </ul> <p>A first blog post describing the main blocks of code that should compose de AD7292 driver. This post will aim to elucidate for the basic structure of an IIO driver comprehensively for newcomers.</p> <ul> <li>Post about AD7292 SPI</li> </ul> <p>Approach how an IIO driver communicates through SPI. What kind of API does IIO have to help to establish communication with an SPI device?</p> <ul> <li>Post about AD7292 operation details</li> </ul> <p>Describe the driver’s probe, read_raw, and write_raw functions. I may want to talk about the IIO channels, the ABI, the support for devicetree and other stuff so this can give origin to a tutorial composed by a series of posts.</p> <ul> <li>Post about Final results on the AD7292</li> </ul> <p>Final post about the operation of the driver as a whole.</p>
The mediapipe-rs project provides a Rust SDK to support mediapipe AI models. The SDK provides utility functions to pre-process application data (such as images, audio and video) into TFLite / PyTorch formats, and convert the inference results back into application data. In order to accomplish this, the [mediapipe-rs](https://github.com/WasmEdge/mediapipe-rs) project has made extensive use of the [ffmpeg](https://www.ffmpeg.org/) library. It [compiles ffmpeg to Wasm](https://github.com/WasmEdge/mediapipe-rs/blob/main/src/build.rs) and then builds it together with the application binary. However, the issue with this approach is that those Wasm-compiled ffmpeg functions are slow. We believe a better approach is to create a ffmpeg plugin for WasmEdge, and allow Wasm applications to call native ffmpeg functions as host functions. - Expected Outcome: - The deliverables will be - A WasmEdge plugin for ffmpeg that is similar to the [WasmEdge OpenCV-mini plugin](https://github.com/WasmEdge/WasmEdge/tree/master/plugins/wasmedge_opencvmini). That is to re-export ffmpeg functions in C style as plugin functions as covered in the [plugin developer guide](https://wasmedge.org/docs/category/wasmedge-plugin-system). - A Rust SDK for Wasm programs to access ffmpeg functions in the plugin. Similar to the [WasmEdge OpenCV-mini SDK](https://github.com/second-state/opencvmini) - Refactor the mediapipe-rs project to use the ffmpeg plugin
For simulating a system that contains many variables and/or equations, traditionally computational engineers tend to focus on utilizing an appropriate method with realistic time and space complexity. The technique of model order reduction (MOR), however, aims at approximating the original model with reduced number of variables and/or equations and still keeping sufficient accuracy. MOR is especially useful in various industries, for example thermal-fluids engineering, micro-electro-mechanical systems and control, where large-scale simulations have to be performed. The methods of model order reduction can be classified into several classes including proper orthogonal decomposition, reduced basis, balanced truncation, etc. Several MOR libraries have been well developed in MATLAB and Python. For instance, pyMOR is a Python library that relies on the application of reduced basis methods to parameterized partial differential equations. MATLAB's model reducer supports pole-zero simplification, balanced truncation mode selection methods and so on. In Julia, the main package that enables symbolic modeling is ModelingToolkit.jl, whose key features consist of automatic transformation and structural simplification, but it needs further works on fast ML approximate transformations. So this project is targeted on implementing MOR methods as extended features for ModelingToolkit.jl. Besides implementation of methods of model order reduction, the expected deliverables will include documentation which shows both the improved simulation time and accuracy penalty for the application of different MOR methods on test problems from various disciplines.
Most scientific and engineering applications heavily depend on floating point arithmetic. Floating point numbers only approximate the real numbers, which can lead to accumulation of errors and ultimately to unusable results. Current state of the art of floating point arithmetic optimization is done in Rust (egg) or Racket (Herbie). This is not very accessible for the actual target group of scientists & engineers who are used to program in high-level languages such as Python, MATLAB and Julia. Additionally, it requires manual labour (rewriting symbolic expressions, then optimizing them, and finally translating the resulting code back to the original language) due to the necessity of interfaces to interoperate between different systems written in different languages. We propose to take advantage of Julia's homoiconicity and powerful existing symbolic-numeric ecosystem to implement an automatic floating point expression optimizer in Julia, acting on any expression that supports TermInterface.jl, such as Symbolics.jl expressions as well as native Julia code. The project will include a package that implements the floating point expression optimization apporach (OptiFloat.jl). To achieve this Metatheory.jl will be fully integrated with Symbolics.jl, enabling e-graph based optimization of any symbolic expression (not just numerical). Additionally we will finalize the release of Metatheory.jl v3.0, identify existing Julia packages that may suffer from numerical error propagation, and perform a thorough comparison of automatically optimized expressions vs. original vs. external system including the time required to obtain them.
This program pairs one or more mentees with an experienced mentor to deliver an AOSP 12 emulator for 64-bit RISC-V. RISC-V International has setup Android SIG and released RISC-V Android github repo as https://github.com/riscv-android-src. Current RVI Android codebase is based on AOSP 10, while the latest AOSP is 12 now. We hope to catch up with latest AOSP 12 codebase, and as part of our plan, an emulator is required first as test and run environment. Deliverables (bullet list of components and the changes expected): - Add RISC-V target in the emulator. - Porting Ranchu virtual board to support RISC-V machine. - Corresponding unit tests implementation - Regression tests pass Note: upon work should be based on AOSP 12. 3. acceptance criteria Acceptance criteria (bullet list with measurable results defined): - Phase I: Virtual peripherals of emulator can work with RISC-V 64bit cores and pass unit tests. - Phase II: Emulator can boot-up with Linux plus basic rootfs w/o GUI for AOSP 12. - Phase III: Emulator can boot-up with Linux plus rootfs w/ GUI for AOSP 12. Note: due to AOSP 12 rootfs is still in developing, a AOSP 10 rootfs from Android SIG repo can be used instead to help verification, but the kernel version should be matched with AOSP 12.
The existing Commons Numbers library for complex types uses collections of complex object instances that incur Java object allocation overhead and is inefficient for iteration and operations on large complex data sets. There are several third party Java libraries for complex math that each have their advantages and disadvantages, but Java developers currently have no single API for Complex operations that can use these libraries at runtime without rewriting their applications This project would provide Java developers with a single Complex API with abstractions that allow for efficient data structures (such as primitive arrays and value types - as proposed by Java Project Valhalla), use of Java 8 language features (Lambdas/Streams) and a Service Provider Interface (SPI) to integrate with existing (JTransforms, JBLAS, EJML) and future third party libraries (Java 17 Vector API, jcuda, aparapi) The project deliverables would include interfaces to represent complex scalar, vector, matrix types and functional interfaces for complex unary and binary operators & functions. The existing methods in Commons Numbers would be refactored as static functions for use as Lambdas with the new API. The JTransforms library will be integrated for supporting operations such as fast Fourier and Discrete Cosine transforms. Optionally, an implementation using Java 17 vector API would be provided for basic operations on complex vectors and matrices. Performance benchmarks comparing different implementations would be published. The deliverables will include complete Pull Requests meeting all Apache Commons Numbers project requirements
The current Wagtail news starter template presents significant onboarding challenges for new developers, including dependency conflicts, Python version incompatibilities, lack of cross-platform setup support, and incomplete features such as a non-functional newsletter and missing RSS feed. This project aims to transform the starter template into a reliable, modern, and beginner-friendly entry point into the Wagtail ecosystem. The approach is structured across three pillars: 1. Stabilization: Introduce CI targeting the latest stable versions, resolve dependency conflicts, enforce reproducible installations, and implement a cross-platform setup system using a single Python entry point validated via CI on both Linux and Windows. 2. Modernization: Replace broken or missing features with Wagtail-native implementations, including a working newsletter (FormPage), RSS feed, improved admin interfaces (PageListingViewSet, SnippetViewSet), SEO fixes, modular StreamField block architecture, and GDPR-compliant asset handling. 3. Developer Experience: Improve onboarding through interactive admin guidance, simplified frontend tooling, better inline documentation, and demo content that works out-of-the-box. Deliverables include a fully functional, CI-validated starter template, modular and reusable architecture, improved documentation, and a setup experience that reduces time from clone to running project to under five minutes. Initial validation experiments show a 100% setup success rate in controlled environments with zero dependency conflicts, demonstrating feasibility.
SMT solvers are widely utilized in computer-guided verification of computer programs and artificial intelligence. With a multitude of theories employed in these solvers, it is beneficial to have access to a diverse range of solvers and to minimize the challenge of learning a new API for each of them. Using the solver’s own API directly makes it difficult to switch to another solver without rewriting extensive parts of the application, as there is no standardized binary API for SMT solvers. Also, since these SMT solvers have varying degrees of capabilities and performance and may not accept the same input, frameworks were created, that offer a unified API to several SMT solvers. JavaSMT has exactly these attributes, a common API for SMT solvers. The framework provides access to a variety of solvers created using Java and other programming languages. Although most solvers share a common set of supported theories and features, their availability of extra theories and performance may differ. Therefore, the inclusion of additional solvers in the framework is advantageous to its users. SMT formulas over the real numbers can encode a wide range of problems in theorem proving and formal verification. Such formulas are very hard to solve when nonlinear functions are involved. For that reason, the SMT solver dReal will be implemented in JavaSMT. The SMT solver dReal is a solver that solves nonlinear formulas over the reals. The solver can handle various nonlinear real functions such as polynomials, trigonometric functions, exponential functions, etc. The goal of the project will be to integrate the SMT solver dReal into the JavaSMT framework and to include it in the existing JavaSMT tests.
RISC-V Performance Modeling SIG is driving the development of Performance Modeling and Simulation Tools for use across RISCV membership to foster collaboration across the community. To that end, the SIG has developed a trace-driven performance model of an example RISC-V superscalar processor using C++ based on the Sparta simulation framework. The performance model, named Olympia, has been adopted by multiple organizations in both industry and academia. We have identified set of features for further development of Olympia: * Add support for execution-driven modeling - https://github.com/riscv-software-src/riscv-perf-model/issues/14 * Model state of art branch predictors along with decoupled frontend using branch-prediction API, with an example being the BOOM processor frontend documented at https://docs.boom-core.org/en/latest/sections/branch-prediction/index.html - https://github.com/riscv-software-src/riscv-perf-model/issues/143 - https://github.com/riscv-software-src/riscv-perf-model/issues/1 * Develop an API for Interconnect model and implement an example Interconnect microarchitecture. - https://github.com/riscv-software-src/riscv-perf-model/issues/60 * Use the API for data/instruction prefetcher and implement an example prefetcher microarchitecture - https://github.com/riscv-software-src/riscv-perf-model/issues/142 * Enhance the existing vector extension implementation Etc.
Debusine currently only shows task logs after a task has fully completed. This means developers working with long-running jobs (such as package builds or test pipelines) have no way to monitor progress in real time or catch failures early. This project adds live log streaming to Debusine. The implementation follows a four-stage pipeline. On the worker side, the sub-process runner will be extended to capture stdout and stderr incrementally as a task runs, sending each line to the server over the existing WebSocket connection rather than waiting for the task to finish. On the server side, a handler will receive these lines and write them to a per-task Redis Stream, providing a persistent, ordered, low-latency buffer that supports multiple simultaneous readers. A server-side view will then expose these log lines to browser clients via Server-Sent Events or a WebSocket endpoint. As a stretch goal, the Debusine web UI will subscribe to that endpoint and display lines appearing on the task page in real time, with no page refresh required. Deliverables: - Worker-side incremental log capture and WebSocket emission, with full test coverage - Server-side Redis Stream ingestion, per-task key namespacing, and post-task flush to PostgreSQL - A server view that streams log lines to browser clients for running tasks - A browser-side UI update that shows live output on the task detail page using plain JavaScript, without any frontend framework
Currently, two major languages are used in high-energy physics (HEP): C++ for numerically intensive code, where execution speed is critical, and Python for interactivity and simplicity of development (frequently used as 'glue' between high-performance code modules). Julia has recently sparked increased attention as a potential language for HEP. This could provide Python's convenient features while maintaining C++'s ideal computational efficiency. In order to continue this investigation, this project will interface the data model library PODIO with Julia. This will allow you to read existing data files into a Julia program. This project seeks to use the same YAML-syntax to auto-generate Julia code for the end user to be utilized in HEP, as well as to do performance testing to compare the language interfaces for C++ and Julia. General Plan PODIO can already : Read YAML files and validate and parse them to extract necessary information like data members, relations and vector members of components and form MemberVariable objects from that information, forming a object dictionary to be used by the jinja2 template engine (using templates for C++) to generate C++ code. Our Plan: Build a prototype without code generation to test whether the current info passed to the jinja2 template Engine by the ClassGenerator class is sufficient, Accordingly adding the pre-processing logic required to the ClassGenerator class create new templates and dictionaries for jinja2 to generate Julia code. Running tests on the Julia code and refactoring generator code. Benchmarking and Documentation.
Deploying a honeypot is only half the battle; the real value lies in the intelligence we can extract from it. Currently, standard honeypot deployments leave valuable attacker logs isolated on individual virtual machines. If an instance is terminated, that threat data is gone forever, making it impossible to correlate global attack patterns or build a cohesive defensive strategy. This proposal shifts the C2SI Honeynet from a collection of isolated servers into a centralized, serverless Threat Intelligence Platform. Using Terraform and Ansible, I will build a modular framework to deploy secure, Dockerized sensor nodes across multiple cloud providers (AWS and GCP). Instead of logging locally, these edge nodes will use Fluent Bit to stream attacker telemetry in real-time to a centralized AWS S3 data lake. From there, an event-driven Python Lambda function will automatically enrich attacker IPs with global threat scores and geolocation data, making the raw attacks instantly queryable for researchers via Amazon Athena. I am not starting this project from scratch. To prove this architecture is viable and cost-effective, I have already engineered the core serverless pipeline, secured the remote Terraform state with DynamoDB locking, and built the hardened GCP modules in my preliminary Pull Requests (PR #8, #13, #14, #15, #35). Over the 350-hour GSoC period, my goal is to scale this proven foundation into a production-ready system that allows C2SI security teams to deploy a global honeynet and hunt threats with a single command.
The project aims to fill the current gap in the creation of a baseline urban grid that allows for network analysis of the morphology of street networks. The idea of the project is to develop a set of tools within the current momepy.preprocessing module that combined with the current integration of NetworkX can output a solid, consistent and repeatable analysis of the urban form. Given that other libraries have also approached this issue (namely osmnx.simplification and citiyseer.tools.graphs), the idea is to contribute from the persisting issues of such libraries to then continue with a solution that is much more independent. Issues like multiple Points on LineStrings, snapping ends, and curving segments have already been addressed by such tools, but parallel LineStrings, and enclosed Linestrings (eg. round-abouts) still need to be resolved, to name a few. It seems that a categorization of the different topological issues (likely similar to the ones just mentioned) could create a framework from which to approach a final pipeline. However, it is proposed that early tests of different potential avenues be performed to sketch a planning route going forward (one of such tests is exemplified under the Approach section). The ideal solution is expected to be source-agnostic, but also leverage the current efforts already done by the community. Overall, the idea is to prioritize work alongside the mentors and the community to focus on increasing the short-term benefits of all PySAL/GeoPandas users (some of whom might already be familiar with some of the other corrections). As the project progresses over different iterative approaches, it is expected to incorporate more performance tests to counterbalance the feasibility of each approach as well as to build the desired autonomy of the tools to be reliant exclusively on input geometries.
Offline change-point detection is a widely used statistical technique for identifying structural breaks in sequential data, with applications in fields such as signal processing, finance, and bioinformatics. Traditional methods, such as prune exact linear time (Killick et al., 2012) and binary segmentation (Hocking, 2024), are unsupervised, modelling shifts in the underlying data distribution without requiring prior knowledge or labels. While effective, they cannot leverage domain knowledge when it is partially available. Weakly supervised change-point detection methods bridge this gap by incorporating small amounts of coarse, inexact, or noisy labels to guide detection. Unlike unsupervised methods, many of which are efficiently implemented with consistent interfaces in packages such as changepoint (Killick et al., 2024) and rupturesRcpp (Nguyen et al., 2025), implementations of weakly supervised approaches are fragmented across packages, with no standardised tooling or shared interface currently available. This project aims to address this gap by extending the mlr3 ecosystem (Lang et al., 2019), a modular and extensible machine learning framework in R that provides unified interfaces for model training, evaluation, and benchmarking, to support weakly supervised change-point detection. It will also establish common abstractions for tasks, learners, and evaluation protocols tailored to change-point problems within the mlr3 design. The resulting framework will be rigorously tested and benchmarked, benefiting practitioners across various fields, enabling systematic comparison across methods, and providing a foundation for future extensions.
<p><a href="https://github.com/0x0f0f0f/Metatheory.jl/" target="_blank">Metatheory.jl</a> is a general purpose metaprogramming and algebraic computation library for the Julia programming language, supporting the novel technique of e-graph term rewriting through equality saturation, first introduced in <a href="https://egraphs-good.github.io/" target="_blank">egg</a>. Our contribution allows performing general purpose symbolic mathematics, manipulation, optimization, synthesis or analysis of syntactically valid Julia expressions and user defined types, with a clean and concise programming interface, both during compilation or execution of programs. We have been currently experimenting with optimizing mathematical code and equational theorem proving strategies. Some of the proposed tasks for this project include:</p> <ul> <li>Experimenting with <a href="https://github.com/JuliaSymbolics/Symbolics.jl" target="_blank">Symbolics.jl</a> integration and optimization for symbolic mathematics.</li> <li>Experiments in optimizing lowered Julia IRs (possibly by statically overlaying methods with <a href="https://github.com/femtomc/Mixtape.jl" target="_blank">Mixtape.jl</a>), allowing users to define concise rewrite systems that directly optimize Julia code.</li> <li>Experimenting in encoding category theory axioms to efficiently model string diagrammatic equational reasoning, and to produce human readable equational proofs. Possibly, integrate this rewrite system with <a href="https://github.com/AlgebraicJulia/Catlab.jl" target="_blank">Catlab.jl</a></li> </ul>
This project will enhance FOSSology's license intelligence stack by revamping both Nirjas and Atarashi into a faster, ML-driven, and language-aware pipeline. Nirjas will move from regex-oriented parsing to a Tree-sitter architecture with language packs across 248 languages, improving comment extraction fidelity in real repositories. Atarashi will be upgraded from broad full-file matching toward retrieval-plus-classification that emphasizes high-signal fragments, because keywords matter for license intent (for example, phrases like permission grant, redistribution, warranty disclaimer, copyleft, and patent terms are often decisive). To support both components, I have already upgraded the Minerva dataset pipeline to generate train/validation/test datasets for Nirjas and Atarashi using merged inputs from ScanCode LicenseDB, FOSSology licenseRef.json endpoint, and The Stack Smol comment corpus, followed by augmentation, LLM-generated hard-negative synthesis, and near-dedup. A sample dataset for benchmarking and integration testing is ready, while the full-scale dataset build is in progress. For modeling, Nirjas will use a model2vec classifier distilled from strong sentence-transformer teachers selected using MTEB/RTEB task performance together with our internal evaluations; the same model2vec distillation path is also planned for Atarashi, with a fine-tuned quantized embedding baseline retained as a fallback depending on speed/accuracy trade-offs. The expected outcome is a production-ready path to more accurate, scalable, and maintainable license detection across diverse codebases.
Stochastic dynamical systems play an important role in scientific and engineering domains, particularly in systems governed by Brownian motion or subject to random perturbations from the environment. Stochastic Differential Equations (SDEs) and Stochastic Partial Differential Equations (SPDEs) provide a mathematical framework to describe such complex stochastic behaviour. Understanding the hidden physics within these systems is pretty much essential for analyzing their stochastic and nonlinear dynamics. The NeuralPDE package in the SciML ecosystem currently lacks documentation for its existing SDE Physics-Informed Neural Network (PINN) solver and does not yet support multidimensional SDEs and SPDEs. This project aims to document the existing solver while also extending it to include PINN solvers for multidimensional SDEs based on the Wiener expansion, which expresses stochastic solutions as spectral expansions in terms of a set of random basis functions. The proposed SDE PINN solver uses the Kosambi-Karhunen-Loève (KKL) expansion for the Wiener process to obtain finite-dimensional approximations, effectively transforming SDEs into rough path ODEs. These ODEs can then be solved using the standard loss function formulation of PINNs for ODEs. As a natural extension, this approach can be generalized to SPDEs using the Polynomial Chaos Expansion (PCE), which extends the Wiener expansion to handle spatially distributed noise in stochastic systems. By developing and implementing these solvers, this project aims to enhance NeuralPDE's capabilities in solving high-dimensional stochastic systems, making it a more comprehensive tool for data-driven modelling and simulation of stochastic dynamical systems.
The BeamVision extension revolutionizes interactive pipeline development in JupyterLab by transforming fragmented notebook visualizations into an integrated workspace, as envisioned in the original design document. At its core lies a dynamic side panel that fundamentally reimagines how users inspect Apache Beam pipelines - consolidating previously scattered visualizations while preserving the documented backward compatibility requirements. This architectural shift directly addresses the project's stated pain points of vertical space consumption and redundant JavaScript code by implementing the described collapsible interface and centralized rendering system. The solution carefully balances innovation with technical constraints through a hybrid approach combining kernel-side HTML generation for reliability with frontend D3.js rendering for interactivity, both explicitly prioritized in the project's technical specifications. Beyond basic visualization, the implementation delivers on the documented cluster management capabilities, providing seamless Flink/Dataproc integration exactly as outlined in the side panel use cases. Every component maintains strict adherence to the original vision - from the multi-tab comparison interface described in the design workflow to the incremental data loading strategy for large datasets. By systematically executing the documented requirements without introducing unsupported features, BeamVision transforms the proposal's vision of "better UX when exploring pipeline states" into a production-ready reality while laying foundations for future extensibility through the approved plugin architecture. The result is a purpose-built environment that finally unlocks Apache Beam's full potential within JupyterLab, precisely as the project architects intended.