Data engineer focused on trustworthy warehouse systems.
Snowflake | dbt | Python | Airflow | Reconciliation | Data Quality
|
|
|
Open-source Reconciliation as Code framework for proving source-target data equivalence. I am building it around the idea that reconciliation should be versioned, readable, repeatable, and useful to engineers before a mismatch becomes an operational problem.
|
Recon Core is shaped around practical warehouse problems:
- source-to-target comparisons
- CDC validation
- schema and row-count checks
- medallion-layer reconciliation
- data quality rules that can live next to code
Beyond TDD: Giving Coding Agents Reusable Regression Memory
Agents, testing, and preserving why regressions matter.
How Uber Turns Data Quality Checks Into a Reliability System
Data quality as reliability, monitoring, and operational response.
What I Learned Building an Agent OS for Agent-Driven Development
Project rules, ADRs, milestone gates, and human-in-the-loop workflows.
model the data clearly
make assumptions explicit
validate the boring cases
document the edge cases
keep the pipeline observable
The best data systems I have worked on were not the ones with the most moving parts. They were the ones where someone could open the model, read the tests, inspect the reconciliation result, and understand what the system believed to be true.
| Area | Tools I reach for |
|---|---|
| Warehouse | Snowflake, SQL Server, PostgreSQL, MySQL, Oracle |
| Modeling | SQL, dimensional modeling, medallion layers, analytics schemas |
| Transformation | dbt Core, incremental models, macros, documentation |
| Orchestration | Apache Airflow, Snowflake Streams, Snowflake Tasks |
| Ingestion | AWS DMS, AWS S3, Snowpipe, Kafka, AWS Glue, PySpark |
| Quality | Reconciliation, CDC checks, schema parity, freshness, precision checks |
| Shipping | Python, Bash, Git, Docker, Kubernetes, ECS/Fargate, Terraform, CI/CD |
- How to make data quality checks useful enough that engineers actually keep them around.
- How to handle no-primary-key tables without pretending the problem is cleaner than it is.
- How to design Snowflake models that stay fast because the physical shape matches the access pattern.
- How to turn reconciliation from a spreadsheet ritual into something reviewable and repeatable.
I came into data engineering after studying statistical physics and simulation-heavy research problems. That still influences how I work: define the model, test the assumptions, make uncertainty visible, and keep the explanation simple enough to survive production pressure.
LinkedIn | Email | Blog | Recon Core