Explore Projects

<p>The current implementation of modeling a loop is quite simple in the Clang Static Analyzer. This simple approach results a loss of coverage, which indicates false negatives. There is already an implemented loop widening method in the Clang Static Analyzer.[1] However, it comes at the price that (almost) all of the MemRegions values will be invalidated. Hence, the false positive rate is relatively high in these paths. That is the reason it is not turned on by default but hidden behind a flag. My project would aim to improve the simulation of the loops in general. Moreover, it would provide an extensible and incremental way of loop widening in which only the relevant regions are invalidated, and thus can be turned on by default.</p> <p>[1] <a href="https://reviews.llvm.org/D12358" target="_blank">https://reviews.llvm.org/D12358</a></p>

The Modeling/etsicatalog project provides VNF/PNF/NS package management service by Micro Service. It also includes a TOSCA parser service which provides generic parser service. As the unified run-time catalog, its VNF/PNF/NS package management interfaces align with SOL003/005 specification. Etsicatalog is a standalone web application based on Python3 and DJango framework. Etsicatalog is going to support SDC Subscription/Notification functions in Guilin version. SDC is the most important design-time component of ONAP. The mentee will be required to make the technical research and take part in the development related this function under the guide of our team. Some work on Alignment with SOL specification for API is also required for the mentee. Besides, We will consider the support for CNF which is a stretch goal (need volunteer）alignment with the CNF task force where applicable.

<p>All electromagnetic phenomena are governed by the Maxwell's equations, which describing how electric and magnetic fields are distributed due to charges and currents, and how they are changing in time. gprMax is open source software that simulates electromagnetic wave propagation by using Yee's algorithm to solve (3+1)D Maxwell’s equations with Finite-Difference Time-Domain (FDTD) method. The behavior of the electromagnetic wave is closely dependent on the material in which it propagates. Some dispersive media have quite complex electromagnetic properties depending on the wavelength. This, for example, means that for different frequencies the wave can propagate with a different speed in different materials. This significantly affects the solver’s output. The main goal of the project is to enhance series of scripts, which modelled electromagnetic properties of the variety range of materials. Initial series of scripts have been prepared, however, their technical improvement and coupling with main software gprMax is required.</p>

<p>The Large Hadron Collider (LHC) at CERN is the world's highest energy particle accelerator, delivering the highest energy proton-proton collisions ever recorded in the laboratory, permitting a detailed exploration of elementary particle physics at the energy frontier. Simulating the particle showers and interactions in the LHC detectors is both time consuming and computationally expensive. Present fast simulation approaches based on non-parametric techniques can improve the speed of the full simulation chain but suffer from lower levels of fidelity. For this reason, alternative methods based on machine learning can provide faster solutions, while maintaining a high level of fidelity. The main goal of a fast simulator is to map the events from the generation level directly to the reconstruction level. That being said, we aim to investigate the efficiency of deep generative models in simulating event reconstructions in a given detector, hence potentially replacing conventional complex algorithms.</p>

<p>Prior to the HL-LHC operation, detector simulations undergo further developments in order to adapt to the increasing amounts of events. Current infrastructure still faces limitations to tackle the expected increase in data in terms of storage capacity and computation time. Early event reconstructions were based on matrix element methods followed by Monte Carlo techniques. Recent advances focused on boosting the speed of event simulations. For instance, C++ - based DELPHES uses simplified detector geometries and particle-material interactions, hence mapping the detector response into a parametric function. Nevertheless, this technique is limited by its exclusiveness to a specific detector geometry at a time, with any detector changes requiring the framework to be adjusted accordingly through hand-coding. Falcon (previously Turbosim), a fast stimulation framework that uses non-parametric methods to discern detector responses without the need for hand-coding. That being said, we aim to investigate the efficiency of deep generative models in simulating event reconstructions in a given detector, hence potentially replacing conventional complex algorithms.</p>

Since the rise of social media two decades ago, researchers have been eager to characterize its impact on society, particularly in the area of information diffusion. Yet, one major gap remains: its reach via television remains unmeasured. As a result, studying the mediated influence of social media has been an ongoing challenge for Social, Behavioral, & Economic (SBE) researchers. Internet Archive’s TV News Archive provides access to over 2.6 million U.S. news broadcasts dating back to 2009. We will use these TV news broadcasts to train object detection and image classification models. “SM LogoTrack” will focus on detecting social media platform logos, while “SM PostTrack” will detect social media post screenshots. By developing robust visual detection models - "SM LogoTrack" and "SM PostTrack" our project will offer researchers and journalists new tools to systematically trace the flow of information from social media to television news. More importantly, this work directly contributes to the mission of the TV News Archive by enhancing the discoverability and analytical value of its vast video collections.

Seeking an escape from daily life chores, many turn to travel as a means of relaxation, renewal, and exploration. However, planning a trip can be overwhelming especially with the abundance of information available online that involves navigating multiple websites, guidebooks and apps to research. Furthermore, discovering the most touristic places in a foreign city or country, along with the best dining spots, shopping locales, and natural vistas, can be a daunting task. The main purpose of this application is to simplify the travel planning experience and inspire users to embark on memorable journeys and discover the most captivating point of interests (POIs) tailored to their preferences. The POIs would be generated by running GEMMA, one of the latest open-source generative text AI models, locally on the AI server in Lleida Lab utilizing Docker technology. Utilizing the liquid galaxy technology, the user would be able to visualize their entire trip on the LG rig with 3 or more screens where the tours along with their info would be sent through KML (keyhole markup language) that will help build beautiful and unique visualizations on the LG rig. In case of unavailability of an LG rig, the user would still have a captivating experience through the app and the integrated google maps. Deliverables - The model trained on the cloud and ready to be inferred in Lleida Lab AI local server. - The new App published on the Play Store under the Liquid Galaxy LAB account. - Full documentation and code on Liquid galaxy lab GitHub.

his project aims to develop AI/ML models for NFV-usecases. Any two of the following three problems can be considered. VNF/CNF resource/performance/failure prediction NFV log analysis with NLP Synthetic monitoring and logging data generation using GANs Learning Objectives ML Techniques: Deep_learning. ML model development AI/ML for Telco Usecases. Interested Interns: DO NOT apply through this platform. Due to a bug, please go to https://wiki.lfnetworking.org/x/EasZB. Read the complete description. When you a ready to apply send your application documents to your Mentor(s) and mentorship@lfnetworking.org

Performance research/design of RISC-V CPU designs require workloads for analysis. Workloads can be custom user applications or industry benchmarks such as SPEC, GeekBench, Dhrystone, etc. Using tools like the RVI Olympia Perf CPU Model users can pinpoint bottlenecks in CPU design, the workload, or the compilers/libraries. When the workload is small, running the workload on Olympia is not complex: trace the workload using a functional model and run that trace through Olympia; count cycles. However, if the workload is large, tracing the entire workload is not practical. Workload reduction tools, such as SimPoint, help narrow down the points of interest (POI) as well as reduce the instruction length to calculate estimated performance. The flow using SimPoint: Workload -> SimPoint analysis (using tools like QEMU) -> workload fragments -> trace generation. Each trace can be run in parallel on Olympia to gather the point performance. Post-processing tools will collate the fragments and generate an overall estimate of performance. This internship will: * Establish workloads (research) * Establish a SimPoint flow to reduce the workloads (QEMU or other instruction set simulators) * Generate STF traces using SimPoint data (QEMU or other instruction set simulators) * Create a repository of traces and their metadata, such as compiler info * Tools to run traces on Olympia and generate perf data (python, C++)

To identify possible relations between gene variants, phenotypes, and diseases - an ideal way would be a multi-model approach that involves: • Supervised learning methods to discover diseases associated with combinations of gene variants and clinical features. • Also unsupervised learning methods like clustering to understand the underlying patterns among diseases that went unnoticed before. This could be a strong evidence-based approach for clinical treatments.

<p>The project aims at generating the Java Model Classes for proposed XML XSD files of repositories. This involves mapping the elements of XML Schema to members of a java class using the tool, which would enable conversion. Conversion is important for the sharing capability of the specifications. Importing packages and customization of tools enabled modification of generated classes. The generated classes would be used as part of a re-designed Java tool for SPDX. This could be used by the organization, whenever required, to map XSD to java classes.</p>

The CTTC 5G-LENA NR module provides a highly detailed and validated simulation of the 5G NR protocol stack within ns-3. However, a critical dimension currently missing is energy consumption modeling. Neither the UE nor the gNB reports power draw or connects to the existing ns-3 Energy Framework. This gap prevents researchers from studying vital metrics like UE battery lifetime, gNB operational expenditure, and overall green networking sustainability. This project will introduce a mature, callback driven energy tracking architecture to the 5G NR module by integrating it with the ns-3 Energy Framework.

OpenCost's data model is now 6 years old, and we have learned a great deal long the way. Based on these learnings, we would like to revamp how OpenCost represents its cost and usage data across the stack. We want to set OpenCost up for the next set of features and scale by building a solid foundation for the way we represent data. Expected Outcome: We would make the UID of Kubernetes objects into a first class citizen. These UID based objects would then be grouped in a similar hierarchy to the way Kubernetes itself is organized. A successful conclusion of this project would see an additional emitter implemented in OpenCost, that emits data objects stored in compressed protobuf which reflect a hierarchy of Kubernetes objects. The mentors will help establish this hierarchy. This hierarchy should then be tested via integration tests against the existing objects, and key metrics compared to be equal across those.

ModelHamiltonians is already a well-structured library to express and simulate Hamiltonians in the context of quantum chemistry. The package, however, lacks some manipulations of the Hamiltonians that can be useful to researchers to efficiently manipulate and analyze electronic structure calculations while maintaining the flexibility to work with different standard theoretical frameworks. To address this, this proposal is related to the creation of a toolkit module that will mainly allow: 1. Atomic Orbital to Molecular Orbital transformations, generalizing to all Hamiltonians are available in the library; 2. Conversions to the geminal basis, enabling researchers to work with alternative representations of Hamiltonians for specialized applications. 3. Reduced Density Matrix (RDM) manipulations, improving the usability of ModelHamiltonians in many-body representations. 4. Add support for FCIDump file format reading and writing 5. Develop an interface with PySCF and write tests for all implemented functionalities.

The goal of the project is to apply and develop “end-to-end” vision transformer (ViT)-based networks for jet-flavor identification with the CMS open data and to benchmark their accuracy performance. This is expected to improve both the accuracy and computational cost of convolutional neural networks (CNN) developed previously for similar tasks. The networks developed would also expand previous works from a binary "light quark vs gluon jets" tagger to multi-class "b quark vs c quark vs light quark vs gluon jets" tagger. The “end-to-end” here refers to a direct use of images recorded by the particle sub-detectors at the CMS for classification. It strips away the complex reconstruction algorithms and object definitions used by the CMS and paths a way to invite ideas outside of the particle physics community to tackle the difficult task of particle identification at the Large Hadron Collider.

The Scrum Helper Chrome Extension project is about improving the way Scrum teams generate daily standup reports by making the process smoother and more efficient. My Google Summer of Code 2025 proposal with FOSSASIA focuses on key upgrades, such as adding support for multiple email providers (Gmail, Yahoo, Outlook), creating a standalone pop-up UI, improving GitHub data fetching, and making the extension faster and more user-friendly. I have already contributed by implementing multi-client email support and will now work on UI refinements, caching, and API improvements. My timeline includes refining the design, improving performance, testing, and writing clear documentation to ensure everything works well across different platforms. With experience contributing to open-source projects like Apache, NumFOCUS, and AI-Alliance, I am eager to continue working with the FOSSASIA community and making this tool more useful for Scrum teams.

This proposal aims to integrate OpenAI's dall·e 3 model into Augmentor's plugin framework, simplifying content creation by allowing the generation of high-quality images directly within Drupal using text prompts. This project will involve developing a user-friendly interface for configuring image generation parameters and previewing results, integrating OpenAI API into Augmentor's module, and documenting the integration along the way.

Medical Procedures in surgery or in interventional radiology are recorded as videos that are used for review, training and quality monitoring. These videos have many features and associated artifacts of which the following are the most used 1. Anatomical Structures i.e organs, tumors, tissues, etc. 2. Medical Equipment i.e scalpel 3. Medical Information overlaid which often contains details that describe the patients conditions. In this project we are going to make use of the video artifacts and train machine learning models using it.

With Sktime’s goal of unifying machine learning with time series, I intend to contribute to that mission with the project titled “Scaling backends, foundation models, PyTorch, darts and pytorch-forecasting with sktime.” Where I’ll mainly be working on the integration with deep learning backends PyTorch through third-party library pytorch-forecasting and time series anomaly detection library darts. With the growth of the usage of deep learning networks in solving time series-related problems, it is needed to integrate sktime with the powerful deep learning libraries, and with this, I am interested in contributing to the Sktime project in this particular subject.

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 "Weighted CRNN architecture" and "CTC based beam-search decoding", this project seeks to develop advanced machine learning techniques capable of accurately transcribing non-standard printed text. The final objective is to create an "end-to-end solution/product" for users, to directly scan a page and obtain its transcription using a smart AI tool.

Automated software engineering helps developers of control software to manage complex industrial applications. Eclipse 4diac has, in recent years, advanced in terms of providing tool support for domain experts. The control software is typically executed in real-time and distributed across devices. Designing such systems is complex due to the need for communication between devices. To detect faults early, modeling implicit assumptions within IEC 61499 software can provide the basis for automated checks. Such assumptions are typically described in constraint languages, which are difficult to use by automation experts. Therefore, this project provides the infrastructure for tool-assisted communication engineering. Where user interactions are required, graphical visualizations are provided. Current infrastructure for generating automated communication does not support the variety of communication protocols that is observed in modern automation systems.

<p>The field of Model Predictive Control (MPC) has seen tremendous progress. The algorithms and high-level software available to solve challenging nonlinear optimal control problems are significantly used in mobile robots to optimize in real-time their path following and navigation. In order to solve Non-Linear Programming Problems (NLP) – which is considered the general form –, we will use CasADi, an open-source tool to solve non-linear optimization problems. Also, the robot should be able to avoid obstacles in real-time by taking into consideration some constraints and penalizing the control values. In practice, previous maps and laser values are combined to create a list of convexified obstacle-free regions. To comply with the real-time requirements of a physical robot, the algorithm will be able to adapt – also in real-time –to the current situation by trading off between execution time and the number of constraints.</p>

Omics data repositories often contain heterogeneous data from multiple studies and diverse sources. This lack of structure in the metadata is challenging for the development of new algorithms and application of machine learning or deep learning to cross-study datasets. Under this project, some work has already been conducted. Currently, Manual review of the metadata schema, consolidation of similar or identical information spread across schema, and incorporation of ontologies where possible has already been done. In this light, manual harmonization cBioPortal’s key clinical metadata across the whole data repository, not just at the study level, and incorporation of ontology terms has improved the AI/ML-readiness of the cBioPortal data. We want to take the current work further to harmonize/digest new/incoming data in the format of the data dictionary already established in an automated fashion with minimum manual curation. For this purpose, we will explore advanced natural language processing techniques, particularly sentence transformers, to automate the process of metadata curation for clinical metadata within the cBioPortal platform.Additionally, we will also creating an interactive dashboard for visualizing and potentially editing the automated harmonization results, enhancing user accessibility and control over the curation process.