Senior Applied Ai Engineer

Description:

Manulife’s Group Functions AI team is scaling AI and advanced analytics capabilities for Actuarial partners to improve how decisions are made and how insights are generated! This role sits within the AI team and focuses on building solutions that use machine learning, optimization, and modern analytical approaches to solve actuarial-adjacent problems at enterprise scale.

In this role, you will take actuarial problems and translate them into AI use cases. These include predictive risk and behavior modeling, grouping, outlier identification, scenario and sensitivity engines, and automation of controls and analytical routines across recurring cycles. The emphasis is on building reusable, production-ready components and analytical products that integrate into business workflows, with clear explainability, strong evaluation, ongoing monitoring, and governance-ready evidence!

Position Responsibilities

You will work closely with actuarial collaborators and engineering partners. Together, you will deliver solutions that are explainable, robust, and operationally balanced. These solutions help accelerate decision cycles, improve consistency, and let teams focus on higher-value judgment where it matters.

Own end-to-end solution design for actuarial AI

Translate actuarial business problems into a clear solution approach: business workflow, data flow, modeling approach, evaluation plan, and operational controls.
Apply strong design thinking: clarify user needs, define decision points, design for adoption, and make trade-offs explicit.
Create lightweight, high-quality design artifacts (e.g., system context, runtime sequence, agent/tool map where applicable, data lineage, decision log) that make build and governance straightforward.
Make smart design trade-offs: accuracy vs explainability, robustness vs speed, and model complexity vs operational sustainability.

Build strong ML, GenAI, and agentic capabilities for actuarial use cases

Develop models such as predictive risk and behavior models, forecasting and scenario models, segmentation, anomaly detection, and optimization approaches.
Build GenAI capabilities such as retrieval-based solutions, structured summarization/extraction, and guided analytical workflows to accelerate insight generation.
Where applicable, design agentic workflows that coordinate multiple steps and tools (e.g., retrieval, calculations, rules, and structured outputs) while maintaining traceability and controls.
Engineer features from large structured and unstructured datasets and ensure solutions remain stable as data and assumptions evolve.

Set a high bar for evaluation and evidence

Define performance expectations with collaborators and implement out-of-time testing, backtesting, error analysis, stability checks, and sensitivity analysis.
For GenAI and agentic workflows, design practical evaluation: scenario coverage, edge cases, human review rubrics, quality scoring, and regression testing.
Document model limitations clearly and build guardrails that ensure outputs are used appropriately.

Partner closely to productionize and operate solutions

Collaborate with data engineering, ML engineering, and software teams to productionize: pipelines, model packaging, CI/CD, deployment, and monitoring.
Implement monitoring for data quality, drift, performance deterioration, and operational failures; define remediation actions when thresholds breach.
Contribute to runbooks and support adoption and UAT with business users.

Work in a governed environment

Produce documentation and evidence required for model risk review, including assumptions, validation results, monitoring plans, and UAT evidence.
Ensure privacy and security expectations are met through data minimization, appropriate access controls, and safe handling of sensitive information.

Raise team capability

Mentor junior scientists through design reviews, code reviews, and evaluation practices.
Help standardize how we build solutions using reusable templates, checklists, and examples to improve consistency and delivery speed.

Required Qualifications

6–10 years of experience in applied data science, machine learning, or advanced analytics, with demonstrated end-to-end delivery into production beyond notebooks, including support for UAT and post-launch iteration.
Strong Python and SQL, with solid software engineering practices: Git-based workflows, code reviews, unit and integration testing, logging, readable code structure, and basic performance tuning.
Hands-on experience with modern DS/ML tooling such as scikit-learn, PyTorch or TensorFlow, and distributed processing platforms such as Spark or Databricks, including feature engineering and model development at scale.
Demonstrated ability to build and communicate solution architecture by producing clear diagrams and short specs. These cover data flow, runtime flow, interfaces, dependencies, failure modes, and operational controls. Align collaborators on trade-offs and scope.
Strong evaluation skills across ML and advanced analytics: backtesting or out-of-time testing, metric selection, error analysis, stability testing, and sensitivity analysis; ability to translate evaluation into business-ready acceptance criteria.
Experience building and operating monitored solutions: data quality checks, drift detection, performance deterioration monitoring, alerting, and practical remediation approaches.
Strong communication and collaborator management: ability to explain outputs, limitations, uncertainty, and build decisions in plain language, and drive adoption in business workflows with domain partners.
Actuarial domain depth demonstrated through significant experience partnering with actuarial teams or solving actuarial-context problems, with comfort in working with actuarial constraints, reconciliation expectations, and governed decision processes.
Working knowledge of GenAI and agentic patterns includes understanding when they add customer value. You should also know how to deploy them responsibly. Experience contributing to a GenAI-enabled capability like retrieval-based solutions, structured summarization/extraction, or tool-using workflows is required.