April 25 ~ 26, 2026, Copenhagen, Denmark
Inés Pérez 1, Lara Suárez 1, Gabriel Novas 1, and Santiago Muíños 1,
Automatic retrieval and standardization of material data from scientific literature remains a critical challenge. This work introduces an artificial intelligence-based framework that extracts and structures material property data and converts it into the International System of Units using state-of-the-art opensource models (Mistral, LLaMA3). The system is highly modular, enabling independent development of specific information retrieval tasks, including PDF parsing, semantic table interpretation, text segmentation, reranking and embedding models for retrieval, and post-processing for data structuring, format validation, and unit conversion. Both supervised and unsupervised evaluation methods are employed to quantify accuracy, consistency, and confidence in the extracted data. By leveraging multiple complementary PDF parsing configurations, we implemented a robust unsupervised evaluation strategy that enhances output reliability. Large Language Models (LLMs) were systematically evaluated using both zero-shot and in-context learning, complemented by a custom scoring system for supervised assessment. The framework was applied to two representative manufacturing use cases (metal additive manufacturing and fiber-reinforced composites) to address the challenge of incomplete, scattered, and heterogeneous material property data in simulation workflows. Results show high extraction accuracy across diverse document layouts, with complementary parsing strategies improving LLM comprehension and tabular processing further boosting overall performance.
Information Retrieval, PDF Parsing, LLMs, Data Standardization, Material Properties
Bin Ge 1, Chunhui He 1, Qingqing Zhao 1, Chong Zhang 1,Jibing Wu 1 1 Laboratory for Big Data and Decision, National University of Defense Technology, Changsha 410073, China 2Unit 96941, Beijing 100085, China
Identifying high-quality frontier topics from massive scientific research data to assist researchers in accurately conducting scientific research is of paramount importance. Traditional analysis methods face bottlenecks such as limited cross-domain adaptability, high resource consumption, and low efficiency. To address these challenges, this study proposes an AI-agent-based frontier topic mining method. An innovative generative-verification dual-agents (D-Agents) architecture is constructed. Specifically, prompt engineering is employed to develop a generative agent (G-Agent), which leverages the semantic understanding capabilities of large-scale pre-trained language models to automatically generate candidate frontier topics. Subsequently, a verification agent (V-Agent) is introduced to establish a multi-dimensional evaluation system, which automatically verifies candidate topics from dimensions including academic novelty, topic accuracy, and completeness of frontier topics. The effectiveness of the proposed method is validated through three manually labeled datasets in computer vision (CV), natural language processing (NLP), and machine learning (ML). Experimental results demonstrate that the D-Agents framework can simultaneously perform frontier topic mining tasks across multiple domains. On the three labeled datasets (CV-DataSet, NLP-DataSet, and ML-DataSet), the D-Agents achieve a precision exceeding 74% while maintaining a recall over 85%. Compared to traditional bibliometric methods, the proposed approach significantly improves precision and recall in frontier topic mining across three distinct fields: altitude sickness, recommendation systems, and oyster reef ecosystems, with performance exceeding 67%. The DAgents framework effectively mitigates the hallucination issue of G-Agent through its automatic generation and self-verification mechanism, thereby substantially enhancing the efficiency of frontier topic mining.
LLMs, Frontier Topics, Prompt Engineering, D-Agents, G-Agent, V-Agent, RAG
Arnold Valaro 1, Lloyd Matak 2
1Independent Researcher
2Lewis-Clark State College, USA
Access to clean and safe drinking water remains a critical public health challenge in rural Malawi, where only 43% of households have reliable access to potable water. This systematic review synthesizes and critically evaluates existing literature on low-cost water purification technologies suitable for rural Malawian contexts, with a focused lens on affordability, sustainability, community adoption, and contextual integration. Traditional methods such as boiling, sedimentation, and chlorination, while culturally embedded, are significantly limited by economic constraints, fuel dependency, logistical hurdles, and persistent knowledge gaps. Emerging decentralized solutions, including bio-sand filters (BSFs), ceramic filters, and adsorbent-based systems, demonstrate considerable technical promise for pathogen and turbidity reduction. However, their long-term effectiveness faces substantial barriers related to material durability, maintenance complexity, supply chain reliability, and socio-cultural acceptance. Recent innovations leveraging locally sourced adsorbents such as rice husk ash, activated charcoal from agricultural waste, and iron oxide-rich red soil offer environmentally sustainable and economically viable pathways for decentralized water treatment. Crucially, this review argues that technological innovation alone is insufficient. Successful and sustainable implementation necessitates integrated, multi-sectoral approaches that synergistically combine appropriate technology design with comprehensive community training, participatory co-design, gender-sensitive planning, supportive local governance, and resilient infrastructure. The review concludes by identifying pivotal research gaps, including the urgent need for robust multi-contaminant removal systems, longitudinal field evaluations under real-world conditions, culturally adapted implementation frameworks, and scalability studies incorporating climate resilience. The findings robustly underscore the indispensable importance of developing and deploying context-sensitive, low-cost, and community-owned water solutions as foundational to achieving Sustainable Development Goal 6 (clean water and sanitation) and improving public health outcomes in Malawi.
Water purification, low-cost technologies, rural Malawi, adsorbents, bio-sand filters, ceramic filters, community engagement, sustainable WASH, appropriate technology, public health.
Youssef Alothman , and Mohamed Bader-El-Den
1University of Portsmouth, UK
2Abdullah Al Salem University, Kuwait
The semiconductor manufacturing industry generates large volumes of highly imbalanced, non-stationary, and operationally critical textual data. Although transformer-based language models achieve strong classification accuracy, their robustness and probability calibration under industrial constraints remain insufficiently addressed, particularly in resource-limited deployments. This paper proposes LiteFormer, a lightweight and calibrated transformer framework for imbalanced industrial text classification. The approach integrates geometry-aware minority oversampling using D-SMOTE, imbalance-sensitive optimization through Focal Loss, and post-hoc temperature scaling for probability calibration. Experimental evaluation on a large-scale industrial Root Cause Analysis corpus demonstrates improved macro-F1 performance and substantially reduced Expected Calibration Error compared to standard transformer baselines, while maintaining computational efficiency. Results under temporal and domain shift further confirm stable performance and reliable confidence estimation.
Imbalanced text classification, lightweight transformers, probability calibration, focal loss, industrial NLP
Yitian Zhao 1, Quincy Stokes 2
1Portola High School, 1001 Cadence, Irvine, CA 92618
2University of California, Irvine, Irvine, CA 92697
This project addresses the lack of accessible, interactive astronomy tools by developing a 3D telescope simulator using Unity [1]. The program models over 9,000 stars with apparent magnitudes above 6, accurately representing their positions, magnitudes, and planetary motions. It integrates three main components: a position plotting system, a rotation management system anchored to Polaris, and a data retrieval system that provides detailed information for each celestial object. Key challenges included scaling stars and planets for visibility, managing rotation dynamics, and organizing data pop-ups for clarity. Experiments tested planet size scaling and star overlap, optimizing visibility while maintaining realism. Methodology comparisons revealed that prior works—Python-based sky maps, ARsupported planetarium apps, and Star Walk 2—offered precision or real-time alignment but faced limitations in interactivity, AR support, device dependence, or star database size [2]. Our project improves usability, database completeness, and offline accessibility, providing an engaging, educational platform while future improvements focus on AR integration and sensor-based orientation.
Telescope Simulator, Universe, Celestial Sphere, Database, Unity
Felix Hirwa Nshuti 1, Krisha Joshi 2
Dr. Pooja Shah 3and Dr. Shakti Mishra 4
Department of Computer Science and Engineering, Pandit Deendayal Energy University Gujarat, India
In the software development landscape, code is written across diverse programming languages, each with unique syntax and semantics. This heterogeneity leads to challenges for cross-language code analysis, optimization, and security assessment. Structural representations of code, addresses these challenge by preserving syntactic and semantic properties for downstream application. This study presents a comparative analysis of representation techniques, including Abstract Syntax Trees (ASTs), Control Flow Graphs (CFGs), Program Dependence Graphs (PDGs), Intermediate Representations (IRs), token-based methods, and vector embedding. Each method is evaluated for tasks such as code classification, summarization, and vulnerability detection. Experimental insights reveal that no single representation consistently outperforms others; instead, hybrid approaches demonstrate improved effectiveness. A cross-language analysis tool leveraging LLVM to extract and visualize IR and CFGs for C, C++, and Python programs is implemented. The findings highlight the necessity of hybrid approaches to for modern multi- language software systems. The implementation can be found at: https://github.com/fnhirwa/struco
Code representation, abstract syntax tree, control flow graph, program dependency graph, code snippets
Haoqun Qin 1, Miguel Angel Cai Lan 2
1 Albany Acedemy, 135 Academy Rd, Albany, NY 12208
2 California State Polytechnic University, Pomona, CA 91768
Ani-Max is a mobile platform designed to prevent health crises in cats and dogs by combining wearable vital sign tracking with environmental risk analysis [1]. Pets can’t voice discomfort, and subtle issues like early dehydration or heat stress often go unnoticed until they escalate. Ani-Max addresses this by continuously monitoring heart rate, blood pressure, activity, temperature, and humidity, using AI to establish a personalized baseline and detect anomalies early. Challenges included maintaining sensor accuracy during movement, ensuring reliable environmental data, and optimizing battery life. In testing, the system performed well at rest but showed accuracy drops during intense activity, and risk scoring declined when weather data was delayed. These insights informed us of improvements like motion filtering and redundant environmental inputs. Ani-Max offers a proactive, real-time health monitoring solution that helps owners act before issues become emergencies, ultimately improving pet well-being and owner peace of mind [2].
Cat and Dog health Monitoring, AI-Enhanced application, Early risk prevention, Vital sign analysis, Real-time health alerts
Jannatul Ferdous Nishi and Mohitul Shafir
Department of Computer Science and Engineering, Independent University,Dhaka-1229, Bangladesh
Setting up appointments is a key part of using resources better and making healthcare systems work better. But patients still miss a lot of appointments, which is a big problem, especially in low- and middle-income countries (LMICs) like Bangladesh, where missed appointments have a bigger effect because there aren’t enough resources. This study introduces a machine learning-based predictive framework for identifying patients likely to miss scheduled appointments by analyzing retrospective healthcare data. Demographic, socioeconomic, behavioral, and environmental variables were employed to develop predictive models, including logistic regression, decision trees, random forests, and gradient boosting. After using SMOTE, the recall for the minority class went from less than 10% to about 30% for models based on trees. This demonstrates that imbalance-aware resampling is effective in healthcare datasets with constrained resources. An examination of feature significance revealed that age, appointment lead time, and scheduling patterns were the predominant influences on attendance behavior. The study highlights the necessity for methodological transparency, the utilization of standardized evaluation metrics (accuracy, precision, recall, and F1-score), and the importance of LMIC-specific contextual factors, such as transportation accessibility and environmental conditions.The findings illustrate that automated statistical prediction systems can enhance scheduling efficiency, help with targeted interventions, and optimize resource allocation in healthcare environments characterized by limited resources, provided they are employed alongside appropriate imbalance management methodologies and robust implementation strategies. This study enhances scalable, data-driven strategies to reduce appointment no-shows in healthcare systems within low- and middle-income countries.
Patient No-Shows, Machine Learning, Healthcare Analytics, Class Imbalance, SMOTE, Appointment Scheduling, Low- and Middle-Income Countries (LMICs), Healthcare Operations.
Xiaotiao Wang 1Tyler Boulom2
1Portola High School, 1001 Cadence, Irvine, CA 92618
2Woodbury University, 7500 N Glenoaks Blvd, Burbank, CA 91504
Tennis is a globally growing sport with many people flocking to it in search of enjoyment or competition, but countless players are ultimately blocked by a financial barrier. A system is proposed to address this issue by providing an alternative to expensive coaching. The project, named MySwingTracker, is comprised of a hardware piece to capture swing motion with an application to display calculated data and connect users to an AI coach that will give feedback based on their data. The hardware was coded in Python and uses a microcontroller, accelerometer, and battery. The application was made using the Flutter Software Development Kit and coded in Dart. When the hardware piece was first tested, it detected extraneous movement such as backswing, so later iterations implemented a calibration system that made the detection ignore certain axes on the 3D plane. The idea of AI coaching without human interaction was tested, and it was validated that the AI system would be accurate and that there would be no decrease in player intensity in training. MySwingTracker can help players of all levels because it focuses on the core of tennis in the swing motion, which is essential for learning tactics and new situational strokes.
Tennis, Accessibility, Hardware/Electronics, AI, App development
Raymond Kaige He 1Moddwyn Andaya2
1Issaquah High School, 700 2nd Ave SE, Issaquah, WA 98027
2California State University, Sacramento, 6000 Jed Smith Dr, Sacramento, CA 95819
This paper addresses the general over-reliance many have on AI in writing, which has been proven to contribute to skill atrophy and reduced creativity [1]. Rather than attempting to replace human authors, this program aims to integrate AI and human writing in a responsible, slightly gamified writing support tool, built upon the foundation of Unity and C#. The system consists of three main components - an AI controller, a task and progress checker, and a main text box where users can write and edit. Instead of generating full essays, the AI attempts to provide the user with creative prompts which it can then check. However, the program encountered several challenges, including bugs, reliable AI validation, and the prevention of user dependency. These challenges required further refinement. Experiments were conducted using twelve AI-generated essays to evaluate whether the system preferentially approves formally structured writing; acceptance rates and evaluation outcomes were quantitatively analyzed. Ultimately, the program aims to represent a practical step toward integrating AI into writing support without replacing human creativity.
Helpful, Assistant, Creative, Support, Artificial Intelligence
Mukiibi Moses
KyungDong University, South Korea
This paper presents a case study on fine-tuning DistilGPT2, a distilled transformer language model with 82 million parameters, for open-domain conversational tasks using the OpenAssistant Conversations (OASST1) dataset. We document the complete experimental pipeline including data preprocessing, model configuration, training dynamics, and qualitative evaluation. The model achieved a validation loss of 1.6963 after two training epochs, demonstrating effective learning of conversational turn-taking patterns. However, generation examples reveal persistent challenges including response repetition and factual inconsistencies, highlighting the limitations of compact language models for complex dialogue generation tasks. This study provides practical insights for researchers working with resource-efficient transformer fine-tuning in conversational AI applications.
DistilGPT2, conversational AI, fine-tuning, transformer models, open-domain dialogue, resource efficient NLP
Hubert Kyeremateng-Boateng, Anthony Herron, Oluwabukunmi David Jaiyeola, Seonho Choi, and Darsana Josyula
Autonomous Technologies Lab, Department of Computer Science, Bowie State University, Maryland, USA
Multi-agent task allocation under strict temporal constraints poses fundamental challenges: agents must coordinate within hard deadlines while managing limited resources and tolerating stochastic failures. We present a hierarchical deep reinforcement learning (HDRL) framework that integrates ex plicit spatiotemporal encoding with a decentralized neural retasking mechanism to address this problem. Specialized neural encoders capture distance-aware spatial relationships and deadline-sensitive temporal dependencies as first-class architectural components, enabling agents to trade off immediate task value against time criticality and heterogeneous priorities. A lightweight retasking network performs real-time reassignment of orphaned tasks upon agent failure without centralized coordination. We evaluate the framework on a multi-agent flower-harvesting benchmark featuring soft and hard deadlines, dynamic priority constraints, and a 10% stochastic agent-failure rate, comparing Double Deep Q-Networks (DDQN) and REINFORCE Policy Gradient (PG) against a Mixed-Integer Linear Programming (MILP) optimum. DDQN consistently outperforms PG across all temporal horizons, achieving near-MILP pollen yield under time pressures of 100–400 timesteps and surpassing the MILP optimum at horizons of 500 timesteps and beyond. These results establish hierarchical spatiotemporal encoding as a viable approach for real-time multi-agent coordination in robotics, logistics, and autonomous systems.
multi-agent reinforcement learning, hierarchical reinforcement learning, spatiotemporal encoding, task allocation, fault tolerance, DDQN
Prof. V.R. Shyanbhag 1
KCS® Dr A.V. Baliga College of Arts and Science, Kumta, Uttara Kannada, Karnataka, INDIA
The concept that actions occur through consciousness is a growing area of research, with experimental findings suggesting that consciousness may exist independently of a functioning brain. This aligns with many seekers’ testimonies, indicating a deeper connection between the m icrocosm and the macrocosm, where personal growth involves perceiving and experiencing untapped potential. The interplay between energy forms is crucial, encompassing both electromagnetic physical waves and psychic waves. These combined waves form what has been termed “Psycho-electromagnetic Waves,” representing a unified interaction of electromagnetic and psychic phenomena. This research investigates the nature of consciousness, exploring its role in bridging quantum mechanics and the classical world, and examining its potential to influence reality through these psycho-electromagnetic interactions.
Microcosm, Electromagnetic, Physical waves, Psychic waves, Quantum mechanics.
Henry Hung Chun Cheng and Alejandro Nava Aldana 1Suffield Academy 2University of California, USA
Volunteer engagement among high school students faces persistent challenges including fragmented resources, lack of accessible training, and difficulty discovering local opportunities. Support Circle is a cross-platform mobile application built with Flutter and Firebase that addresses these challenges by integrating four key components: an AI-powered chat assistant using GPT-4o-mini for real-time volunteer guidance, a Google Maps-based discovery system that executes parallel API searches to identify nearby service opportunities, a structured training module system with reactive real-time progress tracking via RxDart and Cloud Firestore, and a calendar-based service hour logger with atomic batch writes ensuring data consistency. Experimental evaluation demonstrated strong AI response quality averaging 25.8/30 across five volunteer categories, and the parallel search architecture reduced location discovery time by 73.4% compared to sequential execution. By unifying training, tracking, discovery, and AI assistance into a single mobile-native interface, Support Circle provides a comprehensive, friction-reducing platform that empowers young volunteers supporting children and families.
Community Service, Volunteer Management, Mobile Application, Artificial Intelligence, Location-Based Services, Firebase, Flutter, CASA, Service Hour Tracking, Training Modules
Jenay Han1, Teris Han2, Jonathan Thamrun3 1Harvard-Westlake School, 3700 Coldwater Canyon Ave, Los Angeles, CA 91604 2Curtis School, 15871 Mulholland Dr, Los Angeles, CA 90049 3University of California, Irvine, Irvine, CA 92697
Stroke affects approximately 15 million people annually, with survivors facing prolonged physical, cognitive, and emotional challenges during recovery. Communication barriers between patients and healthcare providers often lead to underreported symptoms and diminished care quality. CareLink is an intelligent mobile application built with Flutter that addresses these challenges through five integrated modules: BodyCheck for interactive pain mapping, Symptom Tracking for longitudinal health monitoring, an AI Chatbot for empathetic conversational support, a Communication Assistant with text-to-speech for aphasia patients, and Recovery Activities including music and art therapy. Experimental evaluation of the AI chatbot with 24 diverse stroke patient inputs demonstrated 66.67% first-attempt accuracy, with response quality averaging 4.27 out of 5.0 across six clinical communication criteria. The system excelled in empathetic communication (4.7/5.0 tone score) but showed areas for improvement in recognizing fragmented symptom descriptions as potential stroke indicators. CareLink demonstrates that comprehensive, AI-powered mobile applications can meaningfully support the multifaceted needs of stroke recovery.
Stroke Recovery, Mobile Health, AI Chatbot, Pain Mapping, Mood Tracking, Flutter, Communication Assistance, Patient Care
Tengyue Pan1, Jonathan Sahagun2 1The Stevenson School, 3152 Forest Lake Rd, Pebble Beach, CA 93953 2California State University, Los Angeles, 5151 State University Dr, Los Angeles, CA 90032
Shared laundry facilities in university dormitories present persistent operational inefficiencies, as students lack realtime visibility into machine availability and status. This paper presents LaundrySense, a cross-platform mobile application built with the Flutter framework and Firebase backend services that enables real-time monitoring and management of IoT-enabled laundry devices in student housing environments. The system implements role-based access control separating administrator and student functionalities, leverages Firebase Realtime Database for subsecond sensor data synchronization, and organizes devices through a location-based hierarchy of buildings and rooms. Three major components are analyzed: authentication with role-based routing, real-time IoT sensor streaming via StreamBuilder widgets, and a multi-step device registration workflow with server-side validation. Experimental evaluation demonstrates mean synchronization latencies of 142 milliseconds on campus Wi-Fi, confirming suitability for real-time monitoring. Compared to camera-based, machine-learning-driven, and general-purpose smart campus approaches, LaundrySense provides a focused, low-cost, and privacy-preserving solution for communal laundry management.
Internet of Things (IoT), Mobile Application Development, Flutter Cross-Platform Framework, Firebase Backendas-a-Service, Real-Time Sensor Monitoring
Haoran Zhou1Andrew Park2 1Troy High School, 2200 East Dorothy Ln., Fullerton, CA 92831 2University of California, Irvine, Irvine, CA 92697
DriveWise is a Unity-based driving simulator created to address the problem of unsafe responses to road hazards, especially among inexperienced drivers. Because dangerous situations such as hydroplaning, tire failure, and side collisions are difficult to practice safely in real life, the project proposes a virtual training environment using a Logitech steering wheel and scenario-based simulation [1]. The program links together three main systems: car controls, environmental hazard physics, and interface flow with AI-generated feedback. Important development challenges included creating believable driving physics, collecting meaningful performance data, and generating useful analysis from that data. To evaluate the project, two experiments were designed: one testing whether hazard scenarios were balanced fairly, and another testing whether the AI feedback was helpful to users [2]. The results suggested that scenario tuning strongly affects performance and that feedback quality depends on how specific the recorded data is. Overall, DriveWise offers a safe, interactive, and educational approach to driving practice.
Driving Simulator, Unity, Logitech Wheel, Artificial Intelligence
Marella Likhita Sri Durga, Department of Computer Science, PES University, Bengaluru, India
In this study, we systematically challenge the prevailing assumption that noisy graph construction is the primary performance bottleneck in Graph Neural Network (GNN)-based multi-hop question answering. GNNs were once dominant but have been largely abandoned since 2021, with the information retrieval community converging on the belief that graph noise from named entity recognition errors and coreference failures fundamentally limits performance. Through full-scale reproduction with modern optimization techniques and a novel oracle retrieval graph methodology, we demonstrate that this assumption merits scrutiny. We reproduce the KGNN-RAG architecture on HotpotQA (7,405 examples) and 2WikiMultiHopQA (12,576 examples), achieving 68.6% F1 on HotpotQA—an 18-point improvement over published baselines—without architectural novelty. Critically, our oracle retrieval graph experiments inject perfect ground-truth structure, revealing only 5.0% F1 improvement on HotpotQA and 4.1% on 2WikiMultiHopQA. The consistency of oracle retrieval gaps (difference <1%) across structurally distinct datasets validates our central finding: architectural limitations, not graph quality, are the primary bottleneck for GNN-based multi-hop QA. Error analysis of 1,426 manually categorized failures with high inter-annotator agreement (Cohen’s Kappa = 0.83) confirms this empirically: 64.2% of errors stem from architectural reasoning limitations while only 9.8% stem from graph construction issues. We release complete reproducibility artifacts including oracle retrieval graph datasets, trained models, and comprehensive analysis code, establishing a reliable baseline for future research and demonstrating that careful reimplementation with modern techniques can effectively resurrect dismissed approaches.Our results provide actionable guidance for designing efficient, reproducible multi-hop QA systems in NLP and information retrieval settings.
Multi-hop question answering, Graph neural networks, Reproducibility, Oracle analysis, Architectural bottlenecks.
Weiting Wang1Andrew Park2, California State University, USA
Sudden cardiac arrest claims approximately 350,000 lives annually in the United States, with survival heavily dependent on immediate bystander CPR and rapid AED deployment. Heart Angel is a cross-platform mobile application built with Flutter that addresses this challenge through three integrated components: an accelerometerbased CPR feedback engine providing real-time compression quality assessment via audio and visual cues, a GeoJSON-powered AED locator displaying defibrillator locations on an interactive map, and a Firebase Cloud Functions-driven emergency notification system that dispatches GPS-targeted alerts to nearby CPR-certified community responders. Key challenges including accelerometer noise filtering, audio-visual synchronization, and secure notification delivery were resolved through configurable threshold calibration, timer-synchronized evaluation cycles, and server-side Cloud Function architecture. Experimental evaluation demonstrated 85.3% overall compression classification accuracy across 150 manikin-based trials. Compared to existing standalone CPR training or AED locator applications, Heart Angel provides a unified platform bridging training and real-world emergency response, empowering laypersons to act effectively during cardiac emergencies.
Cardiopulmonary Resuscitation (CPR), Mobile Health (mHealth), Accelerometer-Based Feedback, Automated External Defibrillator (AED) Locator, Emergency Push Notifications
Tianhao Chai1, Bobby Nguyen 1University High School, 4771 Campus Dr, Irvine, CA 92612 4University of California, Riverside, 900 University Ave, Riverside, CA 92521
This project was created to address the challenge swimmers face in managing training, events, personal progress, and technique feedback in one place. Many existing tools focus on only one area, so I proposed a mobile app that combines athlete organization with video-based swim analysis [1]. I built the program using Flutter for the interface and Firebase for authentication, cloud storage, and database management. The app includes profile setup, event tracking, training and nutrition planning, personal records, achievements, and a video upload feature for analysis. One challenge was making different parts of the app work together smoothly while keeping user data stored correctly. I improved this by using a centralized state system and connecting it to Firebase updates. During experimentation, I tested both the accuracy of analysis feedback and the reliability of saved data after reopening the app [2]. The results showed generally strong performance, with some variation caused by video quality and cloud sync timing. This idea is valuable because it gives swimmers one practical, supportive tool.
Swimming, Artificial Intelligence, Pose analysis, Event manager