Data Scientist - 2 (DS-2)

  Job family: Data Science
  Level: DS-2 (equivalent to MLE-2 / AIE-2)
  Scope of impact: Feature
  Theme: Grows and Acts — completes scoped modelling tasks and
  improves team process
  
  Why this role exists

  Product outcomes need deep problem ownership and tight iteration
  with PMs and product engineering. A DS-2 turns a scoped product
  problem into a calibrated model or decision system, ships it
  through the standard production path, and owns its performance
  after launch. You operate with minimal guidance on a defined
  feature, not the whole domain.

  What you own
  
  - A scoped modelling problem framed as a DS task: hypothesis,
  success metric, offline and online evaluation plan.
  - Calibrated predictive or causal models with well-behaved
  probabilities and effect estimates.
  - Repeatable pipelines integrated with production workflows, not
  one-off notebooks.
  - Basic model monitoring for the features you ship.
  - Post-launch performance of your model and its link to the
  target KPI; iterate using telemetry.
  
  What you do not own (yet)

  - Platform uptime and shared serving infrastructure (ML
  Engineering owns this).
  - Domain-wide priority setting across multiple initiatives (DS-3
  and above).

  What you'll do (proficiency expectations at L2)

  Data-driven decision making
  - Build calibrated predictive or causal models with sound
  probability and effect estimates.
  - Articulate the impact of uncertainty and select an applicable
  course of action with minimal guidance.
  - Stress-test findings with simple mental models or simulations
  before trusting them.

  Technical expertise
  - Set up fully reproducible environments for your own work and
  share the guides with peers.
  - Package work into repeatable pipelines and integrate them with
  production workflows.
  - Implement basic model monitoring.

  Applied ML/AI/DS
  - Frame and scope an opportunity as a DS problem, and pick the
  right solution family (prediction, optimization, causal).
  - Review recent literature, build reproducible pipelines, and
  fairly compare alternative models.
  - Run controlled pilots that connect model uplift to a target
  KPI.
  
  Experimentation and inference
  - Frame a testable hypothesis and pick the right design (A/B or
  hold-out).
  - Run multi-metric or stratified tests with power checks and
  CUPED variance reduction.
  - Conclude using confidence intervals, state the limitations,
  and tie results back to a target KPI.
  
  Strategy and influence
  - Scope an opportunity into a well-posed DS problem, naming the
  RoI and the product and process changes it implies.
  - Align stakeholders on the KPI leverage of a proposed approach
  and secure agreement on scope and goals.
  - Coordinate with engineering and product leads to launch
  features where the model provides core value; shape planning and
  risk assessment.
  
  How you work with others

  - PM: co-own the outcome and prioritisation for your feature.
  - Product Engineering: integrate your model into customer-facing
  experiences.
  - ML Engineering / AI Engineering: consume platform primitives;
  collaborate on evaluation, reliability gates, and production
  readiness.

  What we expect from a strong DS-2
  
  - Ships production artifacts on the standard path, not
  prototypes that stall at the production boundary.
  - Improves at least one team process (templates, reviews,
  reproducibility) beyond their own tasks.
  - Owns outcome integrity: model outcomes stay aligned with
  product outcomes after launch.