Projects
See the work of the engineers, researchers and pioneers of ML advancements
Generative AI
Explore the projects that our members have worked on.
Computer Vision and Pattern Recognition
Explore the projects that our members have worked on.
Financial Machine Learning
Explore the projects that our members have worked on.
Medical AI
Explore the projects that our members have worked on.
Supervised Learning
Explore the projects that our members have worked on.
MLOps
Explore the projects that our members have worked on.
AI Applications
Explore the projects that our members have worked on.
See All Projects
Browse all of our AI and ML projects developed by our students
QuDiffuse: Quantum Annealing for Diffusion Models
This project explores the use of D-Wave’s quantum annealing to improve the efficiency of diffusion models, aiming to reduce generation time while maintaining high quality data samples.
Machine Learning in RNA Folding Predictions
Using machine learning to predict RNA folding, this project aims to develop a generative model that maps a given RNA sequence to its 3D molecular structure while optimizing computational energy efficiency.
Deep Emotion: Contrastive Learning for fMRI Classification
This project uses contrastive learning on fMRI brain scans to identify which movie scenes a person is viewing. The goal is to align brain activity with visual stimuli using deep learning techniques.
Flow to Learn: Flow Matching on Neural Network Parameters
This project applies flow matching techniques to neural network parameters, leveraging mixture learning and model weight merging strategies to predict and transfer weights between models.
FashionBot
FashionBot is a personalized clothing recommendation system that intelligently styles complete outfits from catalog data, aiming to provide tailored and stylish fashion recommendations.
AI ChatBot
This AI chatbot answers student inquiries and promotes UTMIST events across Discord and other platforms. It aims to streamline student-club communication with a smart conversational interface.
AI Warehouse: Evolution Battle Simulation
UTMIST is collaborating with AI Warehouse to build an evolution simulation in Unity using C# and genetic algorithms for educational videos reaching millions of viewers.
Lovelytics: Task Automation LLM
Partnering with Lovelytics, this project extends a LangChain-powered LLM automation tool for business workflows, focusing on task generation via natural language prompts on Azure Databricks.
Aercoustics: Sound Event Classification for Construction Sites
This collaboration with Aercoustics focuses on classifying sound events from construction sites using acoustic data, aiding in compliance and environmental monitoring.
Aemulus VR: Using Neural Radiance Fields for Luxury E-commerce
Aemulus VR is using neural rendering and Gaussian Splatting to create photorealistic 3D e-commerce displays for luxury products, transforming online shopping experiences.
KnockRI: AI-Driven Skills Assessment Platform
Knockri uses AI to evaluate candidates through video, audio, and text, reducing hiring bias by focusing on competencies instead of resumes in order to promote fair recruitment practices.
ML Compute Platform Project
This system dynamically manages UTMIST’s cloud-based ML workloads. It handles MLOps workflows, allocates compute resources, logs performance, and provides real-time monitoring across cloud providers.
ML Project Platform
A centralized hub to manage the ML project lifecycle, including task tracking, team collaboration, milestone monitoring, and code/resource sharing for ongoing and past projects.
ECG Analysis Using Deep CNNs
Applying deep convolutional neural networks to analyze ECG data for cardiovascular disease detection.
Employee Attrition Factors Prediction
Utilizing machine learning to identify and address factors contributing to employee attrition.
Real-Time: Real Estate Price Prediction
Developing a real-time model to predict real estate prices using machine learning techniques.
SmileDetector: Detecting and Rating Smiles
Creating a web application that employs machine learning to detect and rate smiles.
WallStreetBots: Stock Price Prediction
Implementing NLP and reinforcement learning for stock price prediction and market analysis.
Handwriting Recognition Using Deep Learning
Applying deep learning techniques to recognize handwritten text in physical documents.
DiffMuse
Exploring music synthesis using diffusion models to generate new compositions.
Numerai Quant Team
Developing models to compete on Numerai's weekly hedge fund prediction challenges.
Virtual Creatures
Reimagining Karl Sims' 1994 'Evolving Virtual Creatures' paper using modern techniques.
12-Lead ECG Reconstruction
Reconstructing full 12-lead ECG signals from only 2-3 lead inputs using ML models.
AltaML - Startup Collaboration
Building an ML pipeline tailored for applications in the construction industry.
FashionBot
Developing a customizable fashion recommendation system using machine learning.
Omni - Startup Collaboration
Integrating information retrieval into the creative process through ML techniques.
Osu!
Generating Osu! beatmaps using sequence-to-sequence models for rhythm game enthusiasts.
PhotoML - Startup Collaboration
Applying machine learning to enhance picture selection processes.
RealTime2
Advancing the RealTime project from 2021 with new features and improvements.
WallStreetBots2
Continuing the WallStreetBots project with a focus on cryptocurrency price prediction.
Wind Turbine Audibility - Aercoustics Collaboration
Providing an ML solution for the turbine audibility problem in collaboration with Aercoustics Ltd.
Side Channel Attacks: Cracking AES with Deep Learning
Crack the AES cryptosystem via side channel attacks using deep learning (ResNet, LSTM, Transformer). The project involves hardware-level data collection and attack modeling.
Auto-Colorization of Historical Black and White Images
Re-implement and improve CNN-based automatic image colorization using GMMs and a specialized architecture, building on the 'Colorful Image Colorization' paper.
Denoising Renderings using Deep Learning
Build a Monte Carlo renderer with deep learning denoising and a real-time GUI viewer. The goal is to improve CGI quality and efficiency.
EMILI: Easy MachIne LearnIng
Construct a dynamic pipeline to collect domain-specific datasets and train LLMs. The project involves web crawling, relevance filtering, and dataset-feedback loops.
Evolving Virtual Creatures... PART II!
Use NEAT and multi-agent RL to evolve and train virtual creatures in Unity, building on last year's successful evolution of walkers and swimmers.
Is Hashing All You Need For Implicit Neural Representations?
Explore why multi-resolution hash embeddings outperform other methods in INRs. Conduct experiments and possibly submit to ICML/NeurIPS 2024.
QuSparse: Quantum Approaches to Sparse Signal Reconstruction
Use quantum algorithms like QFT to improve sparse signal reconstruction, transforming signals into quantum states and reconstructing using QML.
NucleAIse: Protein Function Prediction
Predict protein functions and Gene Ontologies using LLMs (Transformers), GNNs, and CNNs. The project aims to identify key model features and improve biological insights for drug discovery.
Real-Time Trading with Sequence Models and RL
Develop a deep RL-based real-time trading agent for FX and commodities. Design a custom RL environment and agent leveraging sequence models and DRL techniques.
Spiking Neural Networks with Memristor Hardware
Build neuromorphic computing systems using Spiking Neural Networks (SNNs) and memristor-based hardware. Split into software and hardware teams for training and circuit fabrication.