University Hackathon Showcase: The Future of Tech from Collegiate Minds

2024 Hackathons

HackTX

November 2–3

Intel was a proud sponsor of HackTX 2024, held at The University of Texas at Austin (UT Austin) on November 2–3. Participants had the opportunity to explore advanced AI hardware and software technologies from Intel with the Intel® Tiber™ AI Cloud, a unified cloud platform provided as a computing resource during the event. They also discovered how AI PCs from Intel can be used to develop innovative hackathon projects and power future AI workloads.

Best Use of Intel® AI Winners

Out of the 18 hackathon project submissions for the Best Use of Intel AI category, the following projects were winners:

1. First place went to Angel's Protection, a cutting-edge solution that processes live video feeds, storing data based on individuals’ clothing rather than facial recognition, thereby prioritizing privacy. Their system integrated multiple camera feeds that deliver live footage to advanced machine learning models. They employed YOLOv11* (with a custom dataset and trained from the ground up on Intel Tiber AI Cloud) for real-time human detection, coupled with a segmentation model to extract clothing features from detected individuals. The classification then process analyzes the clothing attributes—such as color and sleeve length—ensuring comprehensive data collection. Once the data is processed with the AI PC from Intel, it is securely stored in an InterSystems* vector database. Authorized users can then input natural language queries, which are parsed through an LLM (Llama 3.2) to refine search results before returning the relevant data. All components of the system were powered by Intel, from the model training on Intel Tiber AI Cloud to model inferencing on the AI PC from Intel with Ollama* and other microservices that help it keep up.
2. Second place went to PrimeTime, an advanced productivity app that uses AI to track engagement and optimize time spent on your computer to help you maximize focus and boost efficiency. PrimeTime captures usage data and stores that data in a database to categorize productivity into four categories. From there, PrimeTime shows you the categorizations for each application you use to help you recognize productive and nonproductive behavior. The team used Python*, Typescript, JavaScript*, Redis*, Next.js, NPUs, vectors, LLMs, and OpenVINO™ toolkit. They learned how to use an NPU on an AI PC from Intel for the first time and learned the basics of how to implement OpenVINO toolkit.
3. Third place went to superMAN, which is designed to make command-line usage more intuitive and user-friendly. Intel® Extension for PyTorch* LLMs helped superMAN run faster and more efficiently by optimizing model performance on Intel® hardware. Its tailored optimizations sped up response times and reduced resource usage, allowing superMAN to deliver quick, accurate, and seamless command-line assistance. An added benefit was its integration capability with Ollama, a platform the team was familiar with, which made implementation even more streamlined and versatile. The team used an AI PC from Intel, which enabled superMAN to run AI tasks quickly and securely on local machines. The team states, “With their high processing power, the AI PC from Intel handled intensive AI workloads with low latency, ensuring a responsive user experience. Running locally also enhanced security, as all data stayed on the user's device, protecting privacy and making command-line interactions fast, private, and efficient.”

For more information, see UT Austin’s HackTX 2024 Hackathon: Top Projects Built Using Intel® AI Technologies.

BuzzOnEarth

October 14–15

The Indian Institute of Technology (IIT) Kanpur hosted BuzzOnEarth India Hackathon 2024, India’s largest climate hackathon, on October 14–15. Intel provided participants with high-performance compute resources through Intel Tiber AI Cloud and AI Frameworks and Tools as well as the latest hardware (including CPUs, GPUs, and other accelerators such as the Intel® Gaudi® AI processor). The cloud platform enabled hackathon participants to develop, test, optimize, and deploy high-performance applications using libraries, tools, and frameworks powered by the oneAPI programming paradigm for multiarchitecture, cross-vendor, accelerated computing.

The UrbanEco team from Hansraj College (University of Delhi) won the prize for the best use of AI Frameworks and Tools and Intel Tiber AI Cloud by developing an intelligent EV infrastructure planner project that recommends optimal locations for EV charging stations based on vehicle density, station proximity, and nearby amenities.

For more information, please see Intel-Powered BuzzOnEarth Hackathon Spurs Climate Tech Innovations in India.

DubHacks

October 12–13

Intel sponsored the University of Washington’s DubHacks, offering participants free access to the Intel Tiber AI Cloud platform and AI PCs from Intel to build their hackathon solutions. Attendees had the chance to explore the potential of AI PCs, optimize their projects with oneAPI-powered AI software tools and frameworks, and work with the latest hardware, including the Intel® Gaudi® AI accelerator, available through the Intel Tiber AI Cloud.

Best Use of Intel AI Winners

Out of the 23 hackathon project submissions for the Best Use of Intel AI category, the following projects were winners:

1. SearchMate, the first-place winner, is a desktop extension that runs locally as a smart-search assistant. Users can make a query for information they want from their computer without needing to know the name of the file, such as "find me my biology notes from a few weeks ago on mitosis cell division." It also works with images, so users can find "photo of me and my family in front of statue in Italy" without having to scour their folders and click through each photo to identify it. The team built the back end using Python, specifically working with CLIPModel for embeddings. The front end was built using JavaScript, using the Electron framework to bring a native experience to users all running on an AI PC from Intel.
2. WanOne, the second-place winner, is an LGBTQIA+-friendly AI mental health assistant designed to support users through conversations. The AI uses structured dialogue to build a relationship, check in on the user’s mood, explore deeper emotional concerns, and provide summaries to help users track their mental health over time. The assistant follows a mental health process that’s been fine-tuned for LGBTQIA+ users, offering safety features such as conversation-ending controls and feedback buttons to give users more control over their interaction. The team built WanOne using Intel Tiber AI Cloud and fine-tuned an LLM based on GEMMA 2.
3. brAInstorm, the third-place winner, is a web application that uses generative AI (GenAI) tools to transform writers' messy brainstorming text and audio snippets into coherent summaries and new ideas, enhancing your creative process. brAInstorm allows users to add snippets of text and audio onto a blank whiteboard. They used Whisper to process audio snippets and Perplexity AI* to generate idea summaries and inspiration bits to aid the challenging and unorganized process of brainstorming. The team used AI PCs from Intel to fine-tune Whisper on a dataset of human speech and evaluated the model on reading-related audio snippets. They also quantized the Whisper model to optimize it for faster inferences.

For more information, see DubHacks’24 Hackathon Where Developers Innovatively Utilized Intel Tiber AI Cloud and AI PCs.

MHacks

September 28–29

Intel sponsored the University of Michigan's MHacks, with over 550 participants. Students used AI PC developer kits from Intel and the Intel Tiber AI Cloud to build AI-focused projects. Intel also led two workshops where developers learned to fine-tune a state-of-the-art LLM model in just five minutes using the Intel Gaudi accelerator. The sessions covered the basics of the Intel Tiber AI Cloud, how to make the most of oneAPI tools and code samples, using AI PCs for hackathon projects and future solutions, and building custom GenAI models, among other topics.

Best Use of Intel AI Winners

Out of the 20 hackathon project submissions for the Best Use of Intel AI category, the following projects were winners:

1. AIDJ, the first-place winner, is an AI-powered DJ that enhances live performances by analyzing crowd energy and adjusting music in real time to match the vibe. The team started with a pretrained video-to-text model from OpenVINO. By changing the model’s classification layer to a regression layer, they trained it to output metrics like energy, danceability, and tempo from the video (metrics used by the Spotify* recommendation system). Using these metrics, they were able to match them to the most similar song from a database stored in MongoDB*. Additionally, they use the Spotify API to play the selected track. The team used AI PC developer kits from Intel to create their project.
2. Plaite, the second-place winner, uses computer vision (CV) to recognize and label food items from a webcam, storing this information for future reference. It offers nutritional data from the USDA and features a chat assistant to answer queries and suggest dietary modifications based on user goals. The team used Intel Gaudi software to create this project. They first drew out a basic wireframe of the user interface and listed clear goals for what the application should be able to do. Then, they split up people working on the CV model, user interface, and API calls. While the CV model was still a work in progress, the API calls (USDA and OpenAI) were made to work independently, with to-do segments for future integration. Lastly, when the CV model was ready, it was integrated into the website, with passthrough to the USDA API and OpenAI chatbot.
3. SongChaser, the third-place winner, is a multiplayer game where players navigate between songs by identifying similar tracks based on features defined by Spotify, like danceability and acousticness. They built the base of the program using a K-Nearest Neighbors algorithm trained on Spotify song features with Euclidean distance and filtered by genre. This was built using AI PCs from Intel with a remote desktop.

For more information, see Developers Utilized Intel AI Technologies at the University of Michigan’s MHacks’24 Hackathon.

AI Cal Hacks

June 22–23

Intel collaborated with the University of California, Berkeley to cohost this event. Over 1,000 attendees joined to create AI-focused projects using AI PC developer kits from Intel and the Intel Tiber AI Cloud. Intel hosted workshops and an Intel-focused lab during the event to assist students with their hackathon project.

Best Use of Intel® AI Winners

Out of the 32 hackathon project submissions for the Best Use of Intel AI category, the following projects were winners:

1. Dispatch AI, the first-place winner, is an empathetic agent eliminating 911 wait times during critical emergencies. Dispatch AI fully used the Intel Tiber Developer Cloud for project development and demonstration and extensively integrated Intel AI tools, particularly Intel Extension for Python to optimize their project. Additionally, they implemented a fine-tuned Mistral LLM for specialized emergency response using the Intel Tiber AI Cloud.
2. Batteries by LLM, the second-place winner, predicts the structure of new lithium-ion electrolytes from text description and optimizes using first- principle modeling to fight climate change. Using an Intel® Gaudi® card from the Intel Tiber AI Cloud, they fine-tuned a Llama 2 7B model that converts natural text to modeling input files that reflect atomic positions (POSCAR file).
3. Accel, the third-place winner, is an empathetic chemistry tutor. To build the core capabilities of Accel, the team used Intel Gaudi software and an AI PC from Intel. Intel Gaudi software allowed them to distill a model ('selinas/Accel3') by fine-tuning it with synthetic data they generated from the Llama 70B model to 3B. This enabled them to successfully run their app on the AI PC. The team found the prospect of distributing AI apps with local compute to deliver a cleaner and more secure user experience very exciting, and they also enjoyed thinking about the distributed systems implications of NPUs.

For more information, see Innovative AI Projects.

HackDavis

April 20–21

Intel collaborated with University of California, Davis to cohost this event. Over 800 attendees joined to code for social good. Intel hosted workshops before and during the event to get students started on Intel® technology and assist with their hackathon projects.

Best Use of Intel® Tiber™ AI Cloud Winners

Out of the 11 hackathon project submissions for the Best Use of Intel Tiber AI Cloud category, the following projects were winners:

1. Mad Molecool, the first-place winner, is an all-in-one LLM-powered electronic lab notebook for molecular biologists to easily access common biochemical information. The LLMs were fine-tuned using personally curated datasets with self-annotated data from web-scraped journal articles alongside the cleaner pubmedQA and BiosQA datasets. The main technical stack was a Python* and Flask back end with the LLMs communicating from Intel Tiber AI Cloud through a separate Flask server.
2. Midas Green, the second-place winner, helps democratize access to agricultural technology by identifying plant diseases. Midas Green uses a convolutional neural network (CNN) and fine-tuned Gemma model optimized with int8 quantization from Intel and retrieval augmented generation (RAG) to ensure accurate responses. With the CNN, they can classify 32 plant diseases from images captured in any garden. The back end, hosted on Intel Tiber AI Cloud, efficiently processes the data, while the development environment on Intel Tiber AI Cloud ensures seamless integration and scalability. The model outputs disease classifications, which are then analyzed through a RAG to suggest effective treatment options for farmers.
3. Doggo AI, the third-place winner, is an interactive child’s companion that shows emotional responses for children to have an actual conversation with. The team used Intel Tiber AI Cloud to develop their project for the back end.

For more information, see Innovative AI Projects.

Anohka

April 19

Intel collaborated with Amrita University to host a hackathon at Anohka 2024. Around 1,200 developers joined to develop innovative AI projects using Intel® tools and the Intel Tiber AI Cloud platform. Students also learned from 30 mentoring sessions hosted by Intel.

Best Use of Intel Tiber AI Cloud Winners

Out of the 11 hackathon project submissions, the following projects were winners:

1. Comic-ify, the first-place winner, is an LLM-based application that converts mundane PDFs into visually engaging comic-style content for a better reading experience. Its fine-tuned generative AI (GenAI) model generates imaginative text and images based on text. The team used the Intel Tiber AI Cloud platform and optimized Python* libraries provided by Intel® Distribution for Python*.
    
2. DeepFakeDetective, the second-place winner, is a solution that employs advanced machine learning models to detect and mitigate deepfake media through image classification, video classification, audio analysis, and a combination of these techniques. The AI frameworks used in the project include Intel® Extension for PyTorch*, TensorFlow* optimizations from Intel, and Intel® oneAPI Deep Neural Network Library (oneDNN). The team also used accelerated hardware resources available on Intel Tiber AI Cloud.
    
3. Traffic-signal-AI dynamically adjusts traffic signals to prioritize the passage of emergency vehicles while effectively managing the traffic flow. It uses Intel Extension for PyTorch, Intel® Extension for TensorFlow*, and high-speed CPUs on Intel Tiber AI Cloud for real-time video analysis, audio processing, and optimized machine learning algorithms. AI Tools from Intel help perform faster data processing, model training, and inference in the project.

For more information, see Interesting AI Projects.

LA Hacks

April 19–21

Over 1,000 students joined LA Hacks at the University of California. Students had the chance to learn from Intel experts through multiple workshops based on GenAI, ITRex, and Stable Diffusion*.

Best Use of Intel Tiber AI Cloud Winners

Out of the 16 hackathon project submissions for the Best Use of Intel Tiber AI Cloud category, the following projects were winners:

1. Soundscape, the first-place winner, is an application that generates real-time adaptive music through an AI-driven composition engine. It performs real-time location tracking and user preference analysis to generate soundtracks dynamically. The team used Intel Tiber AI Cloud for back end optimization to ensure responsiveness and scalability of the application. The project was ranked third among the overall LA Hacks winners.
2. Aide-n, the second-place winner, is an application that can help people with common injuries. It uses an image-to-text generation technique to convert the input image of an injury into a textual description, which is then fed into an LLM that generates a potential treatment. The team used Intel® architectures and Intel Extension for PyTorch to build an optimized LLM on Intel Tiber AI Cloud.
3. Scribe, the third-place winner, is an AI tool to empower doctors with data-driven, accurate diagnosis of patients' health conditions. It prevents contradictions such as medication errors and misdiagnoses. The team used the computational power of resources available on Intel Tiber AI Cloud to fine-tune an open source LLM.

For more information, see Innovative AI Projects.

HooHacks

March 23–24

More than 650 students attended the University of Virginia's HooHacks. Competing for a chance to win an AI PC from Lenovo, students used the Intel Tiber AI Cloud to create their projects with the help of Intel experts. Students also had a chance to participate in a fun GenAI activity where they took home a custom tumbler that they built using Stable Diffusion for a Jupyter* Notebook.

Best Use of Intel Tiber AI Cloud Winners

Out of the 14 hackathon project submissions for the Best Use of Intel Tiber AI Cloud category, the following projects were winners:

1. MosaicHealth AI won first place. It is a solution that efficiently filters massive personalized medical data from fitness wearables in real time to help doctors provide critical insights during patient consultations. It helps expedite the diagnosis process by generating comprehensive medical reports. The team extensively used:
   • Intel Tiber AI Cloud to build and deploy its solution
   • Intel® Distribution of Modin* for accelerated data analysis
   • Intel® Extension for PyTorch* for faster inference.
     
     The team also fine-tuned the Phi-2 Transformer-based model using Intel Tiber AI Cloud and used a Prediction Guard (an Intel® Liftoff member) LLM APIs for compliance and reliability.
2. Mind River AI took second place as an innovative AI-based meditation and emotional support assistant that prioritizes user privacy and personalization. The web application creates tailored conversations and meditations for each user, securely storing all interactions and personal information in a private database. Using the LangChain* framework, VectorDB package (powered by Intel’s embeddings), and applying a fine-tuned open source LLM model on a medical dataset on Intel Tiber AI Cloud, MindRiver offers a unique experience enhanced by neuro-inspired art designed to evoke specific emotions.
3. EcoSense secured third place as an AI-powered IoT solution that enhances energy efficiency and sustainability by replacing traditional thermostats with advanced environmental sensors. The system collects data on volatile organic compounds, carbon dioxide, humidity, and temperature to identify human activities using machine learning algorithms. It then adjusts room temperature and air quality dynamically for optimal energy consumption and comfort. The team performed model fine-tuning on Intel Tiber AI Cloud.

For more information, see Innovative AI Projects.

Daksh AI

March 8–10

Intel collaborated with SASTRA University to host a GenAI-focused hackathon at Daksh AI. Students participated in three workshops that focus on GenAI and LLMs that are run by Intel innovators and professors. As a part of their hackathon participation, students accessed the Intel Tiber AI Cloud to build their projects using Intel technology.

Hackathon Winners

Out of the 27 hackathon project submissions, the following projects were winners:

1. Stylist AI, the first-place winner, is a generative AI-based solution for personalized outfit recommendations that can improve user experience on e-commerce websites. The team employed Intel Extension for PyTorch, Intel Extension for TensorFlow, oneDNN, and Intel Tiber AI Cloud.
    
2. Llama Hunt, a second-place winner, is an LLM-based web application that helps find job openings tailored to an applicant's resumes and preferences. The team used Intel Extension for PyTorch for faster inference and Intel Tiber AI Cloud for accelerated deployment of the project.
    
3. Find My Doctor, also a second-prize winner, is a mobile application designed to help users find the best doctors available near them based on symptom analysis and patient and hospital classification, ensuring prompt access to healthcare. The team used oneAPI optimizations of machine learning libraries and Intel Tiber AI Cloud for faster running on GPUs.
    
4. EmergAI, the third-place winner, is a call management solution that enables timely responses to critical situations for efficient emergency services. Using oneAPI optimized machine learning and natural language processing (NLP) frameworks and algorithms, it prioritizes emergency calls in the absence of call handlers by tracking the caller's location and analyzing the call's keywords.

For more information, see Innovative AI Projects.

TreeHacks at Stanford University

February 16–18

In the 10th TreeHacks, 1,500 students participated in Intel workshops, learned from the founder of Prediction Guard, Daniel Whitenack, and explored the Intel Tiber AI Cloud to compete for a chance to win an AI PC from Lenovo.

Best Use of Intel Tiber AI Cloud Winners

Out of the 30 hackathon project submissions for the Best Use of Intel Tiber AI Cloud category, the following projects were winners:

1. Meshworks, the first-place winner, combined mesh-based radio technology with cutting-edge AI processing to build a resilient information network for emergency responders in disaster-struck areas. The project enables connections between different device nodes and the command node in a field, followed by visualizing the summarized reports of fields on a map. The team trained a neural network on Intel Tiber AI Cloud to perform binary image classification to distinguish damaged and undamaged buildings. Intel Extension for PyTorch was used for model optimization, Whisper for speech-to-text conversion, and Prediction Guard LLM for summarization.
2. ScratchML secured second place with a no-code platform designed for teaching machine learning and data analysis principles to students through an easy-to-use GUI and drag-and-drop functionality. The project uses the Prediction Guard LLM API to provide real-time insights into user decisions and outcomes within lessons using Neural-Chat-7B. The team used Intel Tiber AI Cloud to construct and test the sandbox model and to take advantage of PyTorch Optimizations from Intel.
3. Memory Playground achieved the third-place position by creating a web app that boosts memory recall and attention through a memory palace technique (a psychology-based technique of associating mnemonic images in mind to some known places). It helps senior citizens and those with Alzheimer's disease and dementia by generating images of objects that are related to a given setting or environment, as well as related words to the categories. Intel Tiber AI Cloud was used to use the OpenAI API with GPT-4* and the Stable Diffusion* model.

For more information, see Next-Gen AI Developers Take Advantage of Intel Resources at TreeHacks.

Hacklytics at Georgia Institute of Technology

February 9–11

Intel started 2024 by attending Hacklytics. Over 1,200 students had the chance to use Intel Tiber AI Cloud, learn from Intel experts on GenAI and LLMs, and compete for a chance to win an AI PC from Lenovo.

Best Use of Intel Tiber AI Cloud Winners

Out of the 37 hackathon project submissions for the Best Use of Intel Tiber AI Cloud category, the following projects were winners:

1. Robotic Registers won first place with its Graph Generative Adversarial Network (GAN) model designed to synthesize logic circuits using a PyTorch GPU environment on Intel Tiber AI Cloud. The project helps generate datasets crucial for deep learning models in the logic circuit domain. It augments training data for machine learning algorithms, enhancing the robustness and diversity of datasets in electronic design automation. Training models on the generated logic circuit graphs enables advanced anomaly identification within real circuits. Furthermore, the synthetic logic circuits serve as valuable benchmark datasets for testing the performance of various EDA algorithms, facilitating comprehensive exploratory data analysis to unearth patterns, trends, and irregularities in logic circuit design.
    
2. Easy Deep Learning secured the second position in the hackathon. The team developed an innovative, no-code platform to simplify the deep learning model fine-tuning process. The user-friendly platform empowers users to start with just a few images and leverages synthetic data generation through Stable Diffusion and various data augmentations to build and fine-tune models quickly. The unique project architecture integrates a locally developed frontend with Firebase* for real-time data synchronization. It used a Jupyter Notebook with Stable Diffusion on Intel Tiber AI Cloud as the back end engine to craft a sophisticated synthetic data pipeline, seamlessly followed by a custom-written training loop within the same notebook environment.
    
3. Estate Edge, the third-place winner project, uses satellite imagery and deep learning to predict real estate trends by analyzing development over 10 years. It identifies growth areas, advising on smart investments by distance learning. The team used code from one of Intel’s workshops based on wildfire prediction using Intel Extension for PyTorch to get the dataset from Google Maps* mapping service. It used Intel Tiber AI Cloud as the platform for project development and execution.

For more information, see Top AI Projects Built on Intel Tiber AI Cloud at Hacklytics.

Are you interested in having Intel participate at your hackathon? Contact us.

Program Resources

Intel® Student Ambassador Program

Educator Program

Intel Liftoff Program

Intel® Certified Developer Program