Giancarlos Estevez Gian-DS1

`$ whoami`

Data engineer · Santo Domingo, Dominican Republic · open to remote roles

I build data pipelines that hold up outside a notebook: real sources, point-in-time correctness, tests and CI. MSc in Data Science & Business Analytics, currently studying Software Engineering to pair the data side with solid engineering practice.

`$ cat tech-stack.yaml`

`Gian-DS1:~$ cat tech-stack.yaml`
`├─ ◈ data_foundations:` _{Python · SQL · pandas · Parquet}	`├─ ⇄ pipelines_warehouse:` _{Airflow · dbt · Snowflake}
`├─ ✦ machine_learning:` _{scikit-learn · SHAP · FinBERT}	`├─ ▣ api_engineering:` _{FastAPI · Node.js · Express · WebSocket}
`├─ ⚙ delivery_quality:` _{Docker · GitHub Actions · pytest · Vitest · Playwright}	`╰─ ⌁ also_worked_with:` _{AWS · Databricks · Power BI · Streamlit · React}
`focus: data engineering · status: open to remote roles`

`$ ls projects/`

Stock Screener ML

Equity screener for the S&P 500 + NASDAQ 100 built on a strict point-in-time pipeline: fundamentals enter by their real SEC filing date, macro series by their ALFRED vintage, and 8-K sentiment only from the next business day — so the backtest can't learn from the future. Trained on 298k rows (515 tickers, 34 features), with SHAP explanations per signal, a rule-based exit engine and graduated drift monitoring.

Python · scikit-learn · SEC EDGAR · FRED · FinBERT · FastAPI · React · pytest · GitHub Actions

FinTrack · live app

Personal-finance PWA: zero-based budgeting, credit-card cycles, debts and savings goals. Multi-currency, bilingual (es/en), installable, and backed by Postgres row-level security so each user only ever reaches their own rows.

React 19 · Vite · Supabase (Postgres + RLS) · Tailwind · Vitest · Playwright

Zomato Data Pipeline

End-to-end batch pipeline on a modern data stack: ingestion into Snowflake, dbt staging and mart models, Airflow orchestration, and an analytics layer with LLM review enrichment and natural-language querying over the warehouse.

Airflow · dbt · Snowflake · Python · Streamlit

LIENZO

Desktop canvas for running several AI coding agents in parallel — each one in its own real terminal, on an infinite board, driven by text or voice. macOS, Windows and Linux, with no build step.

Node.js · Express · WebSocket · xterm.js · node-pty

`$ python scripts/radar.py`

Coverage across four featured projects: data pipelines 50%, machine learning 25%, web apps 75%, finance 50%, orchestration 25%, agent tooling 25%. Public repository language mix: JavaScript, Python, TypeScript, PLpgSQL, PowerShell and Swift, scaled by relative byte count.

_{Project coverage: share of the four featured projects in each category. Language mix: public repository bytes, square-root scaled relative to the largest language. Snapshot: 2026-10-01.}

`$ git log --stats`

`$ cat background.md`

MSc in Data Science & Business Analytics — IMF Smart Education / UCAV, co-developed with Indra–Minsait
Currently studying Software Engineering
~6 years in customer service across regulated health and finance sectors — real domain knowledge and the habit of absorbing complex business rules fast
Spanish (native) · English (C1)
Certifications: AWS Academy (ML & Cloud Foundations) · IBM Data Science · Google Cybersecurity

`$ connect --socials`

_{Santo Domingo, Dominican Republic · @Gian-DS1 · always learning}
_{City-pop design and SVG generators adapted from macu-dev/macu-dev.}

Navigation Menu