Three projects, end to end: ingestion, orchestration, quality gates, and the business outcome each one moved.
−45% manual prep
Designed Python + SQL ELT on Snowflake and dbt that pulls 8+ source systems into a single analytics layer — 500K+ records a day landing tested, standardized, and ready for Product and Growth.
Data quality: Automated validation rules raised dashboard accuracy from ~88% to 98%
Python · SQL · Snowflake · dbt
Read the case study ↓2 days → 4 hrs
Built reusable Power BI semantic models and DAX measures on top of the analytics layer, so every standard KPI comes from one definition — and teams spin up new dashboards in under a day.
Data quality: Models sit on the validated layer, so 98% dashboard accuracy carries through
Power BI · DAX · Snowflake · dbt
3–4 wks earlier
Built anomaly-detection and risk models over financial and credit data at a capital firm, and optimized the BigQuery warehouse behind them — ~45% faster workloads powering executive BI.
Data quality: Documented lineage and KPI logic — analytics self-service up ~60% in six months
Python · XGBoost · BigQuery · SQL
Data for Product and Growth lived in 8+ source systems. Every report meant hand-pulled exports and hours of prep, dashboards hovered around 88% accuracy, and the weekly reporting cycle swallowed about two days — with rework every month when numbers didn’t line up.
The fix wasn’t another dashboard — it was consolidation: land every source in Snowflake through one ELT path, model it once, and give every team the same numbers. I designed the pipelines in Python and SQL to pull all 8+ systems into a single analytics layer.
Extraction and loading run in Python and SQL; transformations live in dbt, where models are versioned, documented, and reviewed like code. The models standardize KPIs across Product and Growth — 500K+ records a day landing analytics-ready instead of analyst-shaped.
Automated data-quality and validation rules run inside the ELT — so bad data is caught in the pipeline, not on a dashboard. Accuracy climbed from ~88% to 98%, and monthly-report rework dropped 30%.
With Power BI semantic models and DAX measures on top of the trusted layer, the weekly reporting cycle fell from ~2 days to ~4 hours, teams spin up new dashboards in under a day, and manual data-prep is down 45%. The numbers stopped being an argument.
Before: manual pulls · ~88% accuracy · 2-day cycle After: automated ELT · 98% accuracy · 4-hour cycle −45% manual prep
Jun 2025 — Jan 2026
Python · SQL · Snowflake · dbt · Power BI
Sep 2024 — Jun 2025
Python · SQL · ETL
2020 — 2023
Python · BigQuery · SQL · BI
2024
Capstone (AbbVie-sponsored): Python + SQL ETL over 1M+ healthcare records into BigQuery data marts, with XGBoost predictive and anomaly-detection models for executive analytics.
2021
Baku, Azerbaijan.
No proficiency bars — proof instead. Chips link to the project where the tool earned its place.
I’m a data engineer who cares that the number on the dashboard is true. At my last role I took dashboard accuracy from ~88% to 98% — not with heroics, but with automated validation rules built into the pipeline itself.
My path ran from financial-risk analytics in Baku to an M.S. in Business Analytics (STEM) at UC Irvine, and into data engineering in San Diego. The through-line: turning scattered, messy sources into data people can actually act on.
Off hours you’ll find me hiking with a thermos of tea and an audiobook queued.
Open to Data Engineering roles · remote or relocation · available now
I’m actively interviewing and can start immediately. If your team needs data it can trust, let’s talk.
or write to aysan.h@outlook.com
— Aysan