Epistemic governance layer for AI reasoning under uncertainty - ON-1227

Genre de projet: Recherche
Discipline(s) souhaitée(s): Génie - informatique / électrique, Génie, Informatique, Sciences mathématiques, Statistiques / études actuarielles
Entreprise: Digital Placemaking Inc.
Durée du projet: 6 mois à 1 an
Date souhaitée de début: Dès que possible
Langue exigée: Anglais
Emplacement(s): Toronto, ON, Canada
Nombre de postes: 2
Niveau de scolarité désiré: Doctorat
Ouvert aux candidatures de personnes inscrites à un établissement à l’extérieur du Canada: No

Au sujet de l’entreprise: 

Digital Placemaking Inc. is a federally incorporated Canadian deep tech company based in Toronto, focused on advancing civic intelligence and next generation municipal data systems. The company builds KinesisIQ, a behavioral signal architecture designed to help cities understand resident needs through multi modal data ingestion, machine learning–driven pattern detection, and real time civic insights. Digital Placemaking works closely with municipal partners, including a City of Toronto Councillor’s Office, to pilot early warning analytics for emerging issues, service drift, and resident sentiment.
The company’s R&D program extends beyond traditional civic analytics through the development of Aporia, an epistemic governance layer for AI systems. Aporia evaluates signal consistency, contradiction, and underdetermination, enabling safer and more transparent reasoning under uncertainty. This research direction positions Digital Placemaking at the intersection of AI safety, epistemic logic, and applied machine learning.
Digital Placemaking collaborates with academic researchers, including a PhD level AI researcher, and is actively expanding its research partnerships through Mitacs to support experimental development of epistemic assessment, multi branch reasoning, and adaptive inference methods. The company’s mission is to empower cities with trustworthy, interpretable, and resident centric intelligence systems that improve public service delivery and civic decision making.

Veuillez décrire le projet.: 

Digital Placemaking Inc. is developing Aporia, an epistemic governance layer designed to evaluate the reliability, consistency, and uncertainty of machine learning signals used in civic intelligence systems. While KinesisIQ provides multi modal data ingestion and behavioral signal detection for municipal partners, Aporia focuses on the deeper research challenge: enabling AI systems to reason more safely and transparently under conditions of contradiction, drift, and incomplete information.

The goal of this research project is to advance new methods for epistemic assessment, including contradiction detection, underdetermination analysis, multi branch reasoning, and adaptive inference. The company aims to generate new knowledge in the areas of epistemic logic, AI safety, and uncertainty modeling, ultimately contributing to a more trustworthy foundation for civic decision making tools. The final outcome is a validated research framework and prototype reasoning layer that can be integrated into KinesisIQ and shared with academic collaborators.
Interns will work closely with Digital Placemaking’s technical team and academic supervisors to design and test experimental methods for evaluating ML signal reliability. Tasks include: developing algorithms for detecting contradictory or unstable signals; implementing uncertainty quantification techniques; exploring epistemic logic frameworks; constructing synthetic and real world datasets for experimentation; and evaluating reasoning performance under varying levels of noise, drift, and ambiguity. Methodologies may include probabilistic modeling, graph based reasoning, Bayesian inference, multi agent simulation, and formal logic approaches.

This project offers students the opportunity to contribute to a novel research direction at the intersection of machine learning, epistemic reasoning, and civic technology. The work is exploratory, experimental, and intended to produce academically relevant insights, publications, and prototype systems that advance the state of knowledge in AI reasoning under uncertainty.

Expertise ou compétences exigées: 

The ideal candidate will have strong technical foundations in machine learning, probabilistic modeling, or AI reasoning systems, with an interest in epistemic logic, uncertainty modeling, or explainable AI. Required skills include proficiency in Python, experience with PyTorch or TensorFlow, and familiarity with ML workflows such as data preprocessing, model training, and evaluation. Candidates should be comfortable working with structured and unstructured datasets, implementing algorithms, and designing experimental pipelines.

Experience with Bayesian inference, graph based reasoning, probabilistic programming, or uncertainty quantification is an asset. Familiarity with concepts such as contradiction detection, signal drift, anomaly detection, or reliability scoring is beneficial. Knowledge of formal logic, epistemic reasoning, or multi agent systems is a strong plus but not required.

Candidates should be able to read academic literature, design experiments, analyze results, and communicate findings clearly. Strong problem solving skills, curiosity, and comfort with exploratory research are essential. Experience with Jupyter, NumPy, SciPy, pandas, or simulation frameworks is helpful. Prior exposure to civic technology, AI safety, or interpretability research is an asset but not mandatory.