Data Engineer

About Phundit

Phundit is a personal finance platform dedicated to assisting Ghanaians with saving, investing, and accessing credit. Our focus is to develop a secure credit product backed by customers' own savings, which is deterministic, compliant with regulations, and evolves from rule-based to model-driven as more performance data becomes available.

Role Overview

As our inaugural dedicated data engineer, you will establish the essential data infrastructure supporting credit decision processes, producing regulatory documents for our banking partner, and generating labelled datasets crucial for future machine learning engineers.

Primary Responsibilities

• Create a feature store enabling point-in-time queries over 26 defined features.
• Develop transformation pipelines that consolidate event data streams while implementing anti-gaming controls at the source.
• Build daily position files and reconciliation statements for our banking partner.
• Implement data quality monitoring focusing on data freshness, completeness, reconciliation inconsistencies, and detecting distribution shifts.
• Within six months, execute outcome labelling with accurate performance windows and pre-netting status retention.
• Generate the first development sample consisting of one record per matured credit cycle with static features and labelled outcomes.
• Establish monitoring benchmarks, including score distributions, gate failure rates, and Population Stability Index calculations.
• Conduct leakage audits to prevent any post-decision data contamination in feature values.
• Within one year, measure univariate statistics like Information Value and Weight of Evidence for each characteristic.
• Build back-testing frameworks to compare historical decisions, identify discrepancies, and flag deviations.
• Produce vintage performance tables and roll-rate matrices to analyze credit behavior over time.
• Prepare datasets suitable for handoff to machine learning engineers joining the team.

Required Qualifications and Skills

• Advanced knowledge of SQL including window functions, precise numeric types, partitioned tables, and point-in-time self-joins.
• Proficient in Python for scripting, data transformation, and validation.
• Experience with Google Cloud Platform services including warehouse, storage, compute, and task scheduling.
• Skilled in Power BI for building dashboards, regulatory reports, and visualizations of portfolios.
• Solid understanding of event-sourced data models, emphasizing the importance of point-in-time state versus current state in credit decisioning.
• Expertise in precision arithmetic, understanding why floating point representation errors matter in scoring boundaries.

Preferred Expertise

• Background in financial services or credit risk data analytics.
• Experience creating machine learning-ready datasets that consider entity-level data splits, observation and performance windows, and leakage prevention.
• Demonstrated ability to build data infrastructure from the ground up as an early data team member.
• Skills in denormalizing NoSQL document stores into analytical warehouse schemas.

Additional Skills That Add Value

• Familiarity with data transformation tools such as dbt or Dataform.
• Knowledge about Ghanaian financial laws and regulatory policies such as Act 843 and Bank of Ghana reporting requirements.
• Ability to read TypeScript, aiding understanding of emitted backend events though coding is not required.

What We Offer

• A competitive salary reflecting your key role building foundational ML credit data systems.
• End-to-end ownership of the data layer from the onset.
• Clear, detailed technical specifications ensuring purposeful development over vague directions.
• Direct communication and collaboration with the CTO and CEO within a small, flat organizational structure.
• An established Google Cloud Platform environment to work within rather than a greenfield setup.
• Use of Power BI for all dashboarding and visualization linked directly to the data warehouse.
• The unique opportunity to shape a credit system from the first labelled outcome to the deployment of fitted models.