“If you are excited about applying to LLMs to tackle real-world challenges in government procurement, this is a perfect opportunity for you. Be a part of the team of research and machine learning scientists building deployable real-world ML applications from ground up and get mentored by some of the best minds in AI during the process.”
- Maithreye, Machine Learning Scientist
About the Role
This is a paid Residency that will be undertaken over a twelve-month period with the potential to be hired by our client, Publicus, afterwards (note: at the discretion of the client). The Resident will report to an Amii Scientist and regularly consult with the client team to share insights and engage in knowledge transfer activities.
Successful candidates will be members of a cross-functional project team with backgrounds in ML research, project management, software engineering, and new product development. This is a rare opportunity to be mentored by world-class scientists and to develop something truly impactful.
About the Client
Publicus is building the intelligence layer for government procurement, helping businesses identify and win public sector opportunities while enabling governments to improve procurement efficiency, transparency, and decision-making. By transforming fragmented procurement data into structured, actionable intelligence, Publicus streamlines complex procurement workflows and improves trust in government procurement data.
A central objective of this initiative is to move from extracting information out of procurement documents to reasoning over it. Publicus aims to turn the connected record of buyers, vendors, contracts, awards, and amendments into trustworthy, auditable intelligence that supports better matching, sharper competitive insight, and earlier warning of contract risk for businesses and governments alike.
In collaboration with Amii, Publicus is developing next-generation AI capabilities for multi-hop reasoning over the procurement graph, with clear evidence and provenance back to source records. Rigorous evaluation is part of the product: outputs must be auditable so that public-sector users can trust them.
We look for people who always start from the problem. Regardless of any existing solution, the first question is what the actual problem is. Only then do we choose methods based on what current model capabilities and available technology allow. First-principles thinking from the problem is non-negotiable.
About the Project
This project builds on Publicus’ existing document-processing foundation and focuses on the reasoning and evaluation layers above it. Using knowledge-graph methods, entity resolution and linking, graph-based reasoning, Large Language Models (LLMs) where they add clear value on top of graph-based methods and techniques for long-horizon decision-making under uncertainty, the Resident will help develop intelligence capabilities over the graph of buyers, vendors, contracts, awards, and amendments. The goal is not a research artifact but a robust, deployable system whose outputs can be traced and verified. The outputs will serve two groups: vendors and their bid teams (opportunities, competitive insight, and risk signals) and government analysts who need auditable views of the same data.
The project has five core components:
- Entity resolution and linking: reliably identifying the same buyer, vendor, or contract across fragmented records (named entity recognition, disambiguation, and linking), with a confidence score attached to every link.
- Multi-hop reasoning with provenance: graph-based reasoning over the knowledge graph, answering questions that span buyers, vendors, contracts, awards, and amendments, with evidence linked back to the source records and uncertainty-aware reasoning over probabilistic links.
- Matching and competitive intelligence: improving how opportunities, vendors, and competitors are matched and analyzed.
- Early-warning signals on contracts: detecting risks such as upcoming renewals, amendment patterns, and ceiling risk.
- Trustworthy evaluation: designing benchmarks, metrics, and audit methods that cover entity linking, graph reasoning, and final outputs, so that results are verifiable and reliable.
An example problem: Should a vendor start preparing a bid now for a contract that expires in 14 months? The answer depends on the likelihood of renewal, whether the buyer tends to re-tender, and how the contract’s amendments have shifted so far. Acting early costs resources, and acting late loses the opportunity. This is sequential decision-making under uncertainty: the Resident will reason over contract histories that unfold over time, drawing on approaches such as time-to-event models, Bayesian decision theory, bandits, and reinforcement learning.
Required Skills / Expertise
Are you passionate about building great solutions? You’ll be presented with opportunities to both personally and professionally develop as you build your career. We’re looking for a talented and enthusiastic individual with a solid background in machine learning, knowledge graphs, large language models, along with proven experience in applied settings.
Key Responsibilities:
- Design, evaluate, and ship multi-hop reasoning systems over the graph of buyers, vendors, contracts, awards, and amendments, using knowledge-graph and graph-based methods (with retrieval and LLMs where they add clear value on top of graph-based methods), and taking them through to production, not stopping at notebook prototypes.
- Build a reliable graph through entity resolution and linking (named entity recognition, entity disambiguation, record linkage) that connects buyers, vendors, contracts, awards, and amendments across fragmented source records, with calibrated match confidence.
- Build mechanisms that attach clear evidence and provenance to every output, linking conclusions back to the underlying source records, and develop uncertainty-aware reasoning over probabilistic links so that answers reflect the confidence of the links beneath them.
- Develop and improve matching and competitive-intelligence models for businesses seeking public-sector contracts.
- Develop early-warning signals on contracts, such as renewals, amendment patterns, and ceiling risk, using approaches suited to long-horizon decision-making under uncertainty.
- Design the evaluation framework: benchmarks, metrics, and audit methods that make model outputs verifiable, calibrated, and trustworthy, covering entity linking, graph reasoning, and final outputs.
- Conduct applied research in knowledge graphs, graph reasoning, sequential decision-making, and uncertainty estimation, and translate it into robust, deployable systems.
- Ship client-focused MVPs fast and iterate with evidence: tests, benchmarks, logs, and clear verification steps. At Publicus, if it isn't verifiable, it isn't done.
- Engage in regular client meetings, contributing to presentations and reports on project progress.
Required Qualifications:
- A graduate degree (MSc or PhD) in Computer Science or a related field with specialization in machine learning, natural language processing (including entity resolution and linking), graph learning, reinforcement learning or sequential decision-making, or trustworthy AI evaluation or equivalent demonstrated experience shipping applied ML systems. We weigh what you have built and can prove works as heavily as credentials.
- Research or project experience in at least one of: graph-based learning or knowledge graphs, entity resolution and linking, decision-making under uncertainty, LLM reasoning or retrieval, or evaluation of ML systems.
- Proficiency in Python and modern AI frameworks such as PyTorch, Hugging Face Transformers, vLLM, and related machine learning libraries, as well as tools for graph data (e.g., NetworkX, PyTorch Geometric).
- Familiarity with linux, Git version control, and writing clean code.
- A positive attitude towards learning and understanding a new applied domain.
- Must be legally eligible to work in Canada.
Preferred Qualifications:
- Experience with knowledge graphs, graph neural networks, and graph query approaches (e.g., Cypher, SPARQL).
- Experience with named entity recognition, entity linking, or record linkage on messy real-world data.
- Experience with uncertainty-aware reasoning over probabilistic links, or with reinforcement learning, bandits, Bayesian methods, or time-to-event and forecasting models for decision-making under uncertainty.
- Experience with LLM-based retrieval or reasoning over structured data (e.g., graph-based RAG).
- Experience designing benchmarks and evaluation methods for ML systems, including evaluation with partial or incomplete labels,calibration, attribution and provenance, and auditability.
- Experience with deploying machine learning models in production environments or strong software engineering (or MLE) skills is a plus.
- Familiarity with public procurement data or other large-scale transactional and contract data is a plus.
- Publication record in peer-reviewed academic conferences or relevant journals in ML or Applied AI.
Non-Technical Requirements:
- Desire to take ownership of a problem and demonstrated leadership skills
- Interdisciplinary team player enthusiastic about working together to achieve excellence
- Capable of critical and independent thought
- Able to communicate technical concepts clearly and advise on the application of machine intelligence
- Intellectual curiosity and the desire to learn new things, techniques, and technologies
Why You Should Apply
Besides gaining industry experience, additional perks include:
- Work under the mentorship of an Amii Scientist for the duration of the project
- Participate in professional development activities
- Gain access to the Amii community and events
- Get paid for your work (a fair and equitable rate of pay will be negotiated at the time of offer)
- Build your professional network
- The opportunity for an ongoing machine learning role at the client’s organization at the end of the term (at the client’s discretion)
About Amii
One of Canada’s three main institutes for artificial intelligence (AI) and machine learning, our world-renowned researchers drive fundamental and applied research at the University of Alberta (and other academic institutions), training some of the world’s top scientific talent. Our cross-functional teams work collaboratively with Alberta-based businesses and organizations to build AI capacity and translate scientific advancement into industry adoption and economic impact.
How to Apply
If this sounds like the opportunity you've been waiting for, please don’t wait for the closing December 4, 2026 to apply - we’re excited to add a new member to the Amii team for this role, and the posting may come down sooner than the closing date if we find the right candidate before the posting closes! When sending your application, please send your resume and cover letter indicating why you think you'd be a fit for Amii. In your cover letter, please include one professional accomplishment you are most proud of and why.
Applicants must be legally eligible to work in Canada at the time of application.
Amii is an equal opportunity employer and values a diverse workforce. We encourage applications from all qualified individuals without regard to ethnicity, religion, gender identity, sexual orientation, age or disability. Accommodations for disability-related needs throughout the recruitment and selection process are available upon request. Any information provided by you for accommodations will be kept confidential and won’t be used in the selection process.