Analytics Engineer II

Job Description: Remitly is on a mission to transform the lives of immigrants and their families by providing the most trusted financial products and services on the planet. Since 2011, we have been tirelessly delivering on our promises to immigrants sending their hard earned money home. Today, we are proud to have served millions of customers globally with Remitly and our newly launched Passbook app to provide immigrants access to banking. We strive daily to meet our promise to our customers by building peace of mind into everything we do.

About the Role

The Reconciliation Insights and Analytics team builds important financial products and tools that ensure the accuracy and completeness of our financial data. We focus on analyzing and detecting data discrepancies, improving customer experience, and delivering business insight. Our data supports data consumers like Accounting, Treasury, Finance, and Customer Success. You will join a team that owns the platform end to end, data modeling, pipelines, detection, alerting, and the agent workflows that triage and resolve what the platform finds. You will report to the Manager, Data Analytics.

Job Responsibilities

• Own data products end to end — from requirements through modeling, implementation, validation, and production operation
• Design and maintain the metric and check definitions that determine whether financial data is correct and complete, and onboard new business domains onto the platform
• Build and maintain the pipelines and transformations behind those metrics, including dimensional modeling, ETL design, and the documentation and data dictionaries that make them usable by others
• Contribute to the platform’s application and service layer — the interfaces, services and deployments through which partners consume the platform’s output
• Investigate discrepancies and anomalies to root cause, and turn each finding into a durable check rather than a one-off fix
• Use AI agents as a primary working method — to model and reverse-engineer data semantics, to implement and ship production code, and to design the agent workflows the platform itself runs on
• Validate what AI produces: define the acceptance criteria, evaluation, and review that make agent-generated work safe to ship, and apply domain judgment where correctness cannot be automated
• Build alerting and observability so that issues reach a named owner with enough context to act, and so the platform’s own health is visible
• Work with partner teams to understand requirements and ensure data solutions align with Remitly’s goals
• Communicate complex data findings and technical information to non-technical partners in Accounting, Treasury, Finance and Compliance, both verbally and through data visualizations
• Identify ways to improve the reliability, efficiency and quality of the platform, so coverage grows faster than the effort required to operate it

Job Qualifications

• 5+ years as an analytics engineer or equivalent
• Expert in SQL, git, and a programming language (Python, Scala, Java, or similar)
• Experience in dimensional data modeling, ETL design, implementation and maintenance
• Experience with distributed query engines and cloud data platforms (Trino, Spark, Snowflake/Redshift or equivalent) and with workflow orchestration (Azkaban, Airflow or equivalent)
• Knowledge of logging, error handling, and familiarity with CI/CD concepts
• Familiarity with AWS products and services
• Experience with analytical model development and deployment processes
• Strong communication skills and ability to collaborate effectively with stakeholders
• You live our cultural values

AI Proficiency

This role uses AI as a core engineering method, not as an occasional assist. We expect proficiency across three dimensions, and we will ask you to show your work in each:

• Modeling — you use AI to reason about unfamiliar data: reverse-engineering semantics from production tables, proposing metric and check definitions, and stress-testing a data model before you build it.
• Implementation — you ship production code with coding agents as your default method, and can point to work that reached production this way.
• Architecting — you design agentic systems themselves: multi-step workflows, tool and skill design, context and prompt structure, and the evaluation harness that determines whether a workflow is good enough to run unattended.

Most importantly, you can describe how you validate what an agent produces — what you check, what you automate, and where you decide a human must stay in the loop. In a data quality function, being fast is worth nothing if the output cannot be trusted. We care more about this judgment than about any particular tool, mo