Search jobs

Do you want to solve unique challenges specific to the Canadian market? Are you looking to make an impact on a digital commerce landscape through science?

We are hiring an Applied Scientist to join Amazon Canada Technology and push the science boundaries for Canada: the reasoning LLMs, embedding models, cross-encoder rankers, and multi-objective optimization systems that turn billions of products into the right answer for hundreds of millions of Canadian customers. The role spans the full model lifecycle, from understanding and deciphering where the problem lies, building models, mid-training reasoning models on data such as shopping, returns, reasoning, localization to aligning the models with customers on dimensions that matter for shopping through fulfillment. You will be a pioneer on this team and will own defining the long-term Science strategy and execute on the same by partnering closely with several partner teams such as Core Search, Returns and work with CA Product Managers, Software Engineering and BI/DEs.

Key job responsibilities

• Develop personalized multi-modal thinking-LLM techniques that reason about customers, search, queries, ranking and products across both primary and secondary languages.
• Train large language models on various data signals and sets: domain-adaptive continued pre-training, reinforcement learning shopping reasoning traces, instruction tuning for natural-language shopping queries.
• Analyze Canada (CA) ranking-quality gaps using offline metrics and benchmark CA model staleness and architecture against other marketplace (e.g. US, EU, JP) equivalents.
• Formulate falsifiable ranking hypotheses linking a specific feature, architecture, or training change to predicted metric movement.
• Run standardized offline evaluation of model candidates; validate feature statistical distributions and coverage in-locale to detect anomalies; own and lead experiments to drive improvements on offline and online experiment benchmarks.
• Lead error analysis on failing query segments and segment-level deep dives (head/torso/tail, category, device), surfacing CA-specific failure modes such as bilingual and thin-category gaps.
• Own metric-tradeoff analysis and acceptable-regression boundaries, bridging model behavior and CA business priorities.
• Partner with teams such as Core Search on hypotheses, model training, and quality gates (MTBR / WLBR / SSAI).
• Classify feature-expansion effort tiers (available / pipeline extension / locale training / infeasible) combining locale knowledge with science judgment.

About the team

Amazon.ca Technology builds the products and services millions of Canadians rely on every day. Our teams tackle challenges unique to the Canadian market while contributing to Amazon initiatives that span the globe — giving you the best of both worlds: meaningful local impact and the scale of global tech. If you want to help shape Canada's digital commerce landscape within a collaborative Canadian work culture, this is the place to do it. - 3+ years of building models for business application experience