Innovative Projects Realized

Multimodal RAG Explainability

This project “Multimodal RAG Explainability (MRAGE)” aims to develop an interactive tool for explaining Large Multimodal Models (LMMs) augmented with retrieval capabilities. LMMs represent a significant advancement in AI, capable of understanding and generating content across multiple modalities like text and images. By employing the retrieval-augmented generation (RAG) methodology, this project queries external knowledge sources to empower the retrieval capabilities and reduce the hallucinations of the model. Counterfactual reasoning will be utilized to formulate explanations by extracting the input data that directly impacts the answer it generates. This approach will illuminate the decision-making process of multimodal LLMs, enhancing their transparency and interpretability.

The project holds the potential to significantly impact fields such as healthcare, law, and finance, where understanding the rationale behind AI outputs is crucial for trust and responsible deployment. An interactive tool enables users to understand the reasoning behind the model outputs and identify potential biases or errors.

A programmable microalgae cultivation platform for sustainable food production and waste resource recovery

The manipulation of gene activity in microalgae (cyanobacteria) offers the possibility of producing energy and materials directly from sunlight, water, and carbon dioxide, contributing directly to more holistic modes of food production, innovative bioproducts and reliable bioenergy solutions that reduce human carbon emissions. Despite their recognized potential, it has proven challenging to develop scalable genetic systems for industrial cyanobacterial strains. We are addressing this challenge, in collaboration with Purify, a local biotechnology company, using Arthrospira platensis (Spirulina). We will optimize Spirulina growth under a range of metal concentrations using a recently developed high-throughput lighting system at the University of British Columbia. Use of this lighting system will enable us to build a series of models describing optimal growth under a combination of different metal and media conditions, and isolate specific components that are essential for increased biomass yield or metal tolerance and recovery. We will use this combined platform to increase Spirulina carbon capture associated with nutraceutical or food production including overproduction of phycocyanin and selected cofactors with antioxidant or health promoting properties as well as potential applications in water treatment. Throughout this project we will work closely with Purify providing a path to market for the research while training highly qualified personnel primed for success in the emerging Canadian bioeconomy.

Exploring Age-Related Changes in Multisensory Integration using Machine Learning tools

The project “Exploring Age-Related Changes in Multisensory Integration using Machine Learning tools” aims to investigate how changes in neurotransmitter concentrations influence the way young and older adults integrate multisensory information and perceive time. Its main method of investigation involves applying machine learning tools to analyze the extensive dataset collected from behavioral tasks, Magnetic Resonance Spectroscopy, and Transcranial Magnetic Stimulation. This analysis aims to uncover patterns, correlations, and insights.

Optimization of reconfigurable intelligent surfaces placement for non-reciprocal transmission

Reconfigurable intelligent surfaces (RISs) have emerged as a transformative technology in the sixth-generation (6G) wireless communications. The RIS technology can easily be realized in local area networks and disaster management scenarios where ad-hoc local area networks need to be setup to ensure mission-critical connectivity. The objective of this project is to develop the techniques to optimize the number and placement of RF network devices, such as base stations (access points) and transmissive and reflective RISs, in an indoor environment to achieve optimum wireless network performance. This objective can be achieved by assessing the effectiveness of practically feasible RIS-based systems, considering their real-world constraints, and working to bridge the gap between theoretical expectations and actual outcomes. This involves employing creative strategies and methods to simplify complexity, thereby enabling the deployment of more extensive systems. Furthermore, the implementation of an optimized network presents a chance to lower the total power usage of networks, which directly influences the reduction of greenhouse gas emissions.

Evaluating LLMs for Sentence Encoding and Clustering to support Thematic Analysis in Qualitative Research

The project “Evaluating LLMs for Sentence Encoding and Clustering to support Thematic Analysis in Qualitative Research” aims to provide qualitative researchers with advanced machine learning techniques for social media data analysis while maintaining autonomy and ownership of their data analysis. By benchmarking modern Language and Large Language Models against established metrics such as coherence and topic diversity, the toolkit bridges the gap between technical expertise and qualitative research needs. This approach broadens the impact of computational tools across diverse domains like health, education, and governance.
The expected outcome of this project is a report detailing the findings from the benchmarking analysis and user feedback. This report will provide insights into the performance of Large Language Models for thematic analysis tasks and outline recommendations for researchers and practitioners seeking to implement computational tools for qualitative data analysis, contributing to the advancement of knowledge in the field of computational social science.

Improved breeding of spruce in Atlantic Canada: Genetic Strategies for Sustainable Forestry

This research project focuses on enhancing the genetic characteristics of Spruce species in Atlantic Canada. By analyzing data from various sources and reviewing family lineage, we aim to understand better the factors influencing Spruce growth traits. Through this comprehensive approach, we will propose improvements to breeding and deployment strategies, making them more effective in adapting to changing environmental conditions. Our ultimate goal is to develop a specialized Tree Breeding Management System tailored to the needs of stakeholders in Atlantic Canada. This system will empower decision-makers with valuable insights, promoting sustainable practices for managing Spruce forests in the region. Ultimately, our efforts seek to strengthen the genetic diversity of Spruce populations, ensuring their resilience and ecological importance in Atlantic Canada’s forest ecosystems.

Big Data Processing and Analysis

Addictive Mobility is a leading (Big Data) online advertising company in Canada. They use real-time bidding (RTB) platform for online display advertising in mobile devices, where multiple companies compete to show a certain Ad to a specific user at a certain time. The goal is to optimize the system such that they minimize the cost over the campaign period but also send targeted ads to maximize return on investment such as number of clicks or purchases. Massive number of data, which they collect as a result of Ad exchanges, calls for tools such as data visualization, data mining and machine learning methods, which can help to make sense of Big Data. Maximization of return on investment requires learning user behavior from the massive collected data to show the Ad that will maximize the probability of user interaction, and that’s a pure machine learning and optimization problem. Enhancing this process is at the core of the business, and will highly affect the company’s reputation and return.

Software Tools for Discovering Quantum Computing Applications (Differential Equations Focus)

A lot of industries rely on solving complex mathematical problems to improve their processes and design their products. Non-linear differential equations are often essential for making advancements. However, solving these equations with traditional computers is extremely challenging and resource intensive. Current computing methods struggle with these complex equations, leading to high energy use, long development times, and suboptimal design.
One promising avenue for solving these problems is quantum computing which might eventually be able to solve these complex math problems far more efficiently. The challenge is that the development of quantum algorithms and software to support this development is lacking. Solving non-linear differential equations. has received far less attention than applications like quantum simulation. This project aims to build software to develop quantum algorithms for non-linear differential equations. The outcome of this work will be the capability to assess the promise of quantum computing for differential equations solving.

An Update on Stigmatizing Language within the Substance Use Spectrum

Our project addresses a critical issue within the realm of substance use: the pervasive stigma that extends across the entire spectrum of use. By expanding our focus beyond the severe range, which is often the primary focus in research and policy, we aim to identify and understand the stigmatizing language and experiences faced by individuals at various points in the substance use spectrum— including those with no use, beneficial use, and at-risk use. In collaboration with CAPSA, we will conduct inclusive focus groups to gain insights into the stigma individuals encounter and explore actionable strategies for reducing it. By examining the origins of stigma beyond the severe range of substance use, we aspire to uncover its roots and, in turn, support individuals across the entire spectrum of substance use.

Multiphysics Modelling of Low Temperature Urea Injection in Selective Catalytic Reduction Reactor

Based on a thorough literature review in the related research areas, the student will create a 3D COMSOL model to simulate the SCR system for the partner organization. The simulation will be combined with experimental testing on the SCR system, with the goal of optimizing the low exhaust temperature operations of the SCR reactor specifically in reducing the urea deposits formation and the load input from the engine. The partner organization is expected to have benefits including significant savings in experimentation costs, a comprehensive simulation model of the SCR reactors with time-dependent chemical reactions and multiphase flow, and potential product optimization suggestions based on the findings of the project.

Communications intern working within cross-functional teams to develop and commercialize AI-powered solutions in the Public Services sector

AltaML builds artificial intelligence (AI)-enabled solutions to business problems. We work with organisations, bringing together their data and domain expertise with our AI expertise, to develop AI solutions that are deployed in their operations. We also commercialize AI-enabled products business via industry-specific ventures, yielding scalability from our investment in the first solution.
Competition for tech talent is fierce, and our talent strategy includes a talent accelerator program, designed to rapidly equip highly qualified individuals with hands-on work experience in applied AI while providing partners with continuous and cost-effective development of AI solutions. AltaML’s AI Lab for Government, also known as GovLab, is a talent accelerator for public service professionals, post-secondary students and recent graduates. GovLab.ai’s mission is to set a global example of how to transform the public sector through applied AI, and is designed to encourage the growth of technical and business AI skill sets that are in high demand across Alberta and around the world.
The communications intern will be primarily engaged in developing and executing a comprehensive social media strategy plan aimed at showcasing the AltaML team culture, fostering engagement, and highlighting key projects and achievements. They will also take the lead in graphic development and production design initiatives to enhance the visual identity and brand recognition of AltaML & Ventures across various platforms. Additionally, the intern will be responsible for creating compelling and informative content to be disseminated through various channels, including blog posts, newsletters, and social media updates, to effectively communicate the organization’s mission, values, and initiatives. In addressing these activities, the intern aims to tackle challenges such as enhancing brand visibility and recognition in a competitive market, cultivating a cohesive team culture in a distributed work environment, and ensuring consistent engagement and interaction on social media platforms. By executing these tasks, the intern anticipates benefits such as increased brand awareness and recognition among target audiences, strengthened team cohesion and morale, enhanced visual identity and brand consistency, and improved communication and storytelling capabilities, all of which will contribute to the overall success and growth of the partner organization(s). Finally, the intern will play a key role in the planning and development of our event planning as we look to regularly host a series of Innovation Showcase events moving forward. Unlike previous Communications associates (IT 34306) who mainly focused on the development of internal communications, this intern will have a unique opportunity to work with both AltaML and our newly revamped Venture Studio, and will work to both enhance current strategies and develop new ones, depending on the stage of the business/specific internal client they are working with. These newly developed strategies centered around brand awareness, technological innovation, effective storytelling, will play an integral role in the future success and on-going development of our business, especially those on the Venture Studio side as many of them have little to no strategy in terms of Marketing & Communications. For this upcoming term/project, the intern will be focused moreso on the execution and continuous improvement of various larger-scale Marketing and Communications initiatives (GovLab Impact Report, Innovation Showcase, etc), that we anticipate will improve our recognition and awareness in the industry while ensuring we are actively and regularly engaging with the tech ecosystem as a whole.

Design and Implementation of a Type System for Enhanced Mathematical Operations in Rings

The project “Design and Implementation of a Type System for Enhanced Mathematical Operations in Rings” aims to develop a Java library with a straightforward application programming interface (API) that simplifies complex mathematical computations within rings, including solving matrix equations.
While this library is designed as a part for the bigger research project called “Symmetric system for message exchange protocol based on the ring surjection”, its utility extends beyond, serving as a versatile tool for any application requiring efficient, accurate, and memory-effective mathematical computations in rings. The library stands out by providing streamlined error handling and performance optimizations, enabling developers across various fields to integrate sophisticated mathematical functionality into their projects without delving into the complexities of the underlying operations.