Resume Snippets

Data Platform Engineer | Stack Technologies Ltd. | Jun 2025 - Present

Lead engineer on the USA data platform rebuild — replacing a legacy Python ETL stack with a production-grade Dagster-native architecture on AWS, with Polars and DuckDB doing the heavy transformation work. Operating at scale: 1.4B+ rows at the mart layer, 400M+ rows in downstream exposures, ~47GB of parquet across landing/intermediate/marts. Operating as an AI-native developer: custom Claude Code skills, sub-agents and hooks act as force multipliers across scaffolding, data inspection, formatting and review — measurably increasing the rate at which new state pipelines and transformation assets ship.

- Architected production Dagster platform across 15+ US states and 400+ transformation assets (1.4B+ mart rows, 400M+ exposure rows), replacing a legacy Python ETL stack with a schema-contracted, testable modern data platform.
- Cut processing time on critical datasets through targeted Polars/DuckDB rewrites, incremental materializations, and coalesced fact/mart layers.
- Designed reusable exposure/coalesce patterns so legacy mart schemas ride on top of the new fact-table architecture without consumer churn.
- Built a project-specific Claude Code toolchain — sub-agent, slash commands, auto-format hook — that measurably increases the rate of shipping new pipelines.
- Profiled my own Claude Code usage to surface high-frequency patterns (65% parquet inspection) and promoted them into first-class agents/commands.
- Deployed the platform's AWS footprint with Terraform IaC across staging and production, migrating Batch workloads to ECS for efficiency and cost.
- Implemented CI/CD (GitHub Actions, ruff, pyright, asset schema-contract tests) and the branching/deployment automation that makes staging → prod boring.
- Standardized 100+ Dagster assets with reusable framework patterns and wrote the docs so the team scales beyond me.
- Raised the engineering bar: schema contracts, type-safe Python, architectural docs, and review workflows befitting an enterprise-grade data platform.

Modern Data Platform Architecture:
- Architected production Dagster platform across 15+ US states and 400+ transformation assets (1.4B+ mart rows, 400M+ exposure rows), replacing a legacy Python ETL stack with a schema-contracted, testable modern data platform.
- Cut processing time on critical datasets through targeted Polars/DuckDB rewrites, incremental materializations, and coalesced fact/mart layers.
- Designed reusable exposure/coalesce patterns so legacy mart schemas ride on top of the new fact-table architecture without consumer churn.
AI-native Developer Workflow:
- Built a project-specific Claude Code toolchain — sub-agent, slash commands, auto-format hook — that measurably increases the rate of shipping new pipelines.
- Profiled my own Claude Code usage to surface high-frequency patterns (65% parquet inspection) and promoted them into first-class agents/commands.
Cloud Infrastructure & DevOps:
- Deployed the platform's AWS footprint with Terraform IaC across staging and production, migrating Batch workloads to ECS for efficiency and cost.
- Implemented CI/CD (GitHub Actions, ruff, pyright, asset schema-contract tests) and the branching/deployment automation that makes staging → prod boring.
Framework & Team Enablement:
- Standardized 100+ Dagster assets with reusable framework patterns and wrote the docs so the team scales beyond me.
- Raised the engineering bar: schema contracts, type-safe Python, architectural docs, and review workflows befitting an enterprise-grade data platform.

Lead engineer on the USA data platform rebuild — replacing a legacy Python ETL stack with a production-grade Dagster-native architecture on AWS, with Polars and DuckDB doing the heavy transformation work. Operating at scale: 1.4B+ rows at the mart layer, 400M+ rows in downstream exposures, ~47GB of parquet across landing/intermediate/marts. Operating as an AI-native developer: custom Claude Code skills, sub-agents and hooks act as force multipliers across scaffolding, data inspection, formatting and review — measurably increasing the rate at which new state pipelines and transformation assets ship.

- Architected production Dagster platform across 15+ US states and 400+ transformation assets (1.4B+ mart rows, 400M+ exposure rows), replacing a legacy Python ETL stack with a schema-contracted, testable modern data platform.
- Cut processing time on critical datasets through targeted Polars/DuckDB rewrites, incremental materializations, and coalesced fact/mart layers.
- Designed reusable exposure/coalesce patterns so legacy mart schemas ride on top of the new fact-table architecture without consumer churn.
- Built a project-specific Claude Code toolchain — sub-agent, slash commands, auto-format hook — that measurably increases the rate of shipping new pipelines.
- Profiled my own Claude Code usage to surface high-frequency patterns (65% parquet inspection) and promoted them into first-class agents/commands.
- Deployed the platform's AWS footprint with Terraform IaC across staging and production, migrating Batch workloads to ECS for efficiency and cost.
- Implemented CI/CD (GitHub Actions, ruff, pyright, asset schema-contract tests) and the branching/deployment automation that makes staging → prod boring.
- Standardized 100+ Dagster assets with reusable framework patterns and wrote the docs so the team scales beyond me.
- Raised the engineering bar: schema contracts, type-safe Python, architectural docs, and review workflows befitting an enterprise-grade data platform.