Lead Data Scientist job description

Data and analytics — Lead Data Scientist
Sets modeling and experimentation direction across a team.

RESPONSIBILITIES
- Rework a transformation layer that has accumulated silent assumptions, replacing them with tested, documented logic.
- Re-architect a warehouse area that has become slow or expensive, and quantify the improvement before and after.
- Untangle the lineage of a messy legacy dataset nobody has documented, and design how lineage is captured so it stays accurate as pipelines change.
- Design access for data of mixed sensitivity across several systems, and set the review for new data uses covering bias and appropriateness.
- Take a model, including a pre-trained AI model, to production with tests on real cases and monitoring for drift.
- Forecast through a launch, promotion or structural shift, and state what the model cannot predict before the forecast drives a commitment.

REQUIREMENTS
- Experience at the Senior Data Analyst level (level 3 of 5) or equivalent
- Holds AWS Certified Data Engineer - Associate and Salesforce Certified Tableau Desktop Foundations

REQUIRED SKILLS
Data Engineering
- Analytical SQL Querying — You can diagnose and rewrite a query that is timing out a production dashboard, and you set the SQL style and review standards your team follows.
- Dimensional Data Modeling — You can redesign a model that has drifted from the business it describes, migrating its consumers without breaking their reports.
- Data Pipeline Development — You can design the pipeline architecture for a new data domain, including how failures are surfaced and how a bad run is safely reversed.
- ETL/ELT Transformation Design — You can rework a transformation layer that has accumulated silent assumptions, replacing them with tested, documented logic.
- Data Warehouse Architecture — You can re-architect a warehouse area that has become slow or expensive, and quantify the improvement before and after.

Data Governance
- Data Quality and Pipeline Monitoring — You can trace a failure through a chain of dependent pipelines to its source, and set which checks block a run and which only warn.
- Metric Definition and Governance — You can set the metric governance process for a domain, including how a new metric gets approved and where its single source of truth lives.
- Data Lineage and Documentation — You can untangle the lineage of a messy legacy dataset nobody has documented, and design how lineage is captured so it stays accurate as pipelines change. You get other teams to document what they own.
- Data Ethics and Privacy — You can design access for data of mixed sensitivity across several systems, and set the review for new data uses covering bias and appropriateness.

Statistics & ML
- Statistical Analysis and Inference — You can design the statistical approach for a complex or contested analysis, and catch a misuse of statistics in someone else's work before it ships.
- Experiment Design and A/B Testing — You can set the experimentation standards for a product area, including guardrail metrics, and stop a launch being pushed through on a cherry-picked result.
- Predictive Modeling and Machine Learning — You can take a model, including a pre-trained AI model, to production with tests on real cases and monitoring for drift.
- Feature Engineering — You can decide which features a model needs from the data available, test whether they actually improve it, and drop the ones that do not earn their place. You keep them interpretable enough to explain to the business.
- Forecasting and Trend Analysis — You can forecast through a launch, promotion or structural shift, and state what the model cannot predict before the forecast drives a commitment.
- Data Exploration and Preparation — You can work out what the fields in a poorly documented dataset mean from the data itself, and build preparation that reruns to the same result.
- Programming for Analysis — You can review shared analysis code and improve its tests, structure and documentation so other analysts can reuse it.

Visualization
- Dashboard Design — You can set the dashboard standards for a domain, covering performance, shared definitions and accessibility, and retire dashboards nobody uses.
- Data Visualization Best Practice — You can present awkward data, such as many series, wide ranges or heavy uncertainty, so it stays readable and accurate. You review what others publish and can say precisely what a design overstates.
- Stakeholder Analytics Translation — You can agree a project plan with senior stakeholders, covering scope, milestones and what the data can and cannot tell them.
- Analytics Storytelling and Reporting — You can take a difficult or unwelcome finding to senior stakeholders and keep the argument standing under challenge. You coach others to cut the detail that interests an analyst but changes no decision.

Delivery
- Project Management — You can keep several people's work in step across teams, agree priorities and hand-offs with them, and keep one shared task list that all of them actually use.
- Planning & Estimation — You can estimate work that spans several teams, name the assumptions behind each figure, and rework the plan when those assumptions break.
- Ownership & Accountability — You can hold accountability for outcomes delivered mostly by other people, taking the blame when it fails and passing on the credit when it works.
- Quality Focus — You can set how quality is checked on complex work shared by several teams, and you spot the problems that only appear where their pieces meet.

Craft
- Problem Solving — You can make the call on a problem outside your expertise and explain its cost, time and risk.
- Domain Expertise — You can bring in good practice from outside the organization and make it work here, and your judgment clearly improves the work you touch.
- Continuous Learning — You can judge which new ideas are worth the team's time and which are not, and you make room for the people around you to learn too.

Communication
- Communication — You can bring disagreeing groups to agreement on the facts and the next steps, and help colleagues frame a difficult or sensitive message.
- Collaboration — You can align teams with competing priorities on a common goal, surfacing the conflict early instead of letting it harden into resentment.
- Writing & Documentation — You can own the documents behind a large piece of work, coordinating contributions so that people who were not involved can carry it on.
- Stakeholder Management — You can hold senior and external relationships, negotiate between competing demands, and deliver unwelcome news without losing trust.

Leadership
- Leadership — You can lead across team boundaries, build the credibility that makes people follow you by choice, and create room for others to lead.
- Mentoring — You can develop other mentors, coach people through career decisions rather than tasks, and lift the capability of a whole team.
- Strategic Thinking — You can set direction for an area, choose deliberately what not to do, and defend that choice when it is challenged.

REQUIRED CERTIFICATIONS
- Salesforce Certified Tableau Desktop Foundations (required from Data Analyst)
- AWS Certified Data Engineer - Associate (required from Senior Data Analyst)