Data Scientist – Pilot Machine Learning

Role overview

CoAd is hiring a Data Scientist – Pilot Machine Learning to conceive, prototype, and validate new machine learning models using CoAd’s internal data. This role focuses on the front end of the ML lifecycle—problem identification, statistical/methodological validation, and producing handoff-ready artifacts for production by the Staff MLOps Engineer.

Responsibilities

• Model ideation & problem framing
  • Identify prediction/estimation problems solvable from existing internal data assets.
  • Review operational data for signal and scope problems with business stakeholders.
  • Write a pre-registration defining target variable, success criteria, and identification strategy.
  • Obtain sign-off from the AI Experimentation Lead before starting prototyping.
  • Aim for at least two novel model proposals per quarter, supported by data feasibility checks.
• Statistical design & experimental validation
  • Run exploratory analysis, feature relevance tests, and baseline benchmarks.
  • Use controlled or quasi-experimental designs when relevant.
  • Serve as methodological author of record, documenting identification strategy, assumptions, and known failure modes prior to MLOps handoff.
  • Apply methods including:
    • Supervised learning baselines
    • Time-series decomposition
    • Causal inference (e.g., difference-in-differences, propensity matching, synthetic control, interrupted time-series)
    • Power analysis for experiment sizing
• Prototype development & handoff packaging
  • Build prototype model code suitable for MLOps handoff (not notebook-only).
  • Produce handoff artifacts including:
    • Versioned repository
    • Documented training pipeline
    • Reproducible validation results
    • Model card
    • Handoff brief (serving contract, retrain frequency, monitoring schema, data dependencies)
• Collaboration with the data team on feasibility
  • Validate data feasibility before adding models to the backlog (availability, refresh cadence, data quality, tenant isolation, governance constraints).
  • Provide early pushback on infeasible proposals.
• Documentation & transparency
  • Ensure methodological transparency with complete model cards for models reaching handoff stage.

Requirements

• Experience spanning multiple ML/statistical methods (not limited to a single toolkit).
• Ability to frame prediction/estimation problems and define identification strategies and success criteria.
• Strong capability to conduct exploratory analysis, benchmark models, and design/evaluate causal or quasi-experimental approaches.
• Ability to produce production-oriented prototype code and documentation suitable for handoff.

Nice-to-haves

• Familiarity with experiment sizing (power analysis) and end-to-end ML lifecycle handoffs.
• Experience working closely with MLOps and data teams to validate data feasibility and governance constraints.