Skip to content
Profile · Finance & DataAs of September 2026

Davirson Novoa Ramírez

Finance Data Analyst

Target roles: Financial BI Analyst · Analytics Engineer

I read a P&L, and I build the pipeline that feeds it.

Economist and FP&A consultant supporting operations across 15+ countries. I run a production data platform — daily ingestion, automated quality tests, a Power BI model — that I built and operate myself.

Verifiable figures

Live pipeline · refreshes daily

Every figure links to what proves it.

Figure · Financial-inclusion index by department

Two maps of Colombia by department showing the composite financial-inclusion index on one diverging scale: in 2018 the country reads in neutral tones with a red periphery, and by 2025 almost the whole territory is blue.20182025
Below the 2018 baseline · 2018 baseline · Above the baseline

In 2018, seventeen departments sat below the baseline. In 2025, one does.

Both maps share one scale, and the index is standardised against 2018. The whole country moved up at once — and that common rise is exactly why the effect vanishes: with entity effects alone, financial inclusion predicts growth (+0.0242, p < 0.001); take the year out and the coefficient is zero.

Open the atlas of all 1,123 municipalities

What I built, and the problem it solves

One project, told the way a case is told: the problem first.

credit-risk-mlops

Decision system with model governance

The problem

A credit model nobody can audit does not get deployed, however good it is. The validator does not ask what the AUC is: they ask who can change that figure without anyone noticing, what happens when the regime shifts, and how you know the model still sees the population it was trained for.

What I built

A credit decisioning system over 1.96M SBA 7(a) loans and 62.4M HMDA applications, with out-of-time validation across the COVID shock, ten gates that block promotion of a model that does not comply, a model card and validation report generated from the run, drift monitoring, and a causal inference layer. No published figure is written by hand: the gate recomputes them from the saved predictions before letting anything be promoted.

Why it matters

This is what separates a model from a deployable one. The same structure — thresholds derived and written down, documentation that regenerates itself, controls that fail closed — is what a model risk examination asks for, and what makes a number survive the question of where it came from.

PythonLightGBMPyTorchDuckDBPySparkMLflowONNXPower BI

Published finding

My first AUC was 0.9461 and I deleted it: it was a leak. And of the ten gates, one blocks my own access model at a disparate impact ratio of 0.7639 against a 0.80 threshold — I did not move the threshold.

Also on the desk

  1. Financial inclusion and regional growth in ColombiaReproducible research · open data

    Nineteen public sources in a dimensional warehouse on dbt and DuckDB, every series resolved to municipal codes. On top: a financial-inclusion index by dimension, two annual panels, an atlas of all 1,123 municipalities and a full econometric battery, with its results published.

    BUILDING
  2. JARVIS — personal tracking appOwn product · Next.js + Supabase

    Daily logging of habits, body, sleep, food and spending on Postgres with a row policy on 34 tables, 526 tests and eight CI gates. There is an open demo running the five real screens on data from someone who does not exist: it never queries the database, and a test enforces that.

    Private repository · the demo is the public surface

    OPEN DEMO
  3. market-data-medallion — data platformProduction platform · refreshes itself

    Daily ingestion from four market sources into a PostgreSQL warehouse in medallion layers with dbt, 89 automated quality tests and CI/CD, on free infrastructure. Of more than 1,300 strategy variants evaluated on top of it, barely one in eight of the in-sample winners survived out of sample — I published every one that did not.

    IN OPERATION
  4. Seven tables in TMDL over the warehouse, 17 DAX measures and four report pages, versioned as text in the public repository. The full catalogue, expression by expression, is on its page.

    IN THE REPO

Track record

See full CV
  1. Business Consultant, FP&A · Neoris EPAM

    2026 — present

    Financial management systems for North America: close, forecast and SG&A variance across 12 countries. Fully remote.

    FP&A

  2. SLB · intern to specialist in 26 months

    2024 — 2026

    LATAM treasury and billing: FX analysis in Python, automation that freed ~10 hours a month per analyst — about 60 across the team —, revenue recognition under SOX.

    Finance + data

  3. Economic research · LEE Javeriana

    2023

    Applied research and volunteer social analytics with international remote teams.

    Data

Tools, with the proof next to them

Each tool with the work that backs it. Everything here is running today, not sitting on a certificate.

Excel and financial modelling
SG&A close and forecast across 12 countries at Neoris EPAM
SQL · PostgreSQL
Three-layer medallion warehouse, more than 58,000 candles in production
Python
Incremental ingestion, backtesting engine, FX decomposition
dbt
89 quality tests that run before a single figure is published
Git · GitHub Actions
Daily cron in operation, with a rate-limit circuit breaker
Machine learning
Stanford Specialization on Coursera, 2024

Hiring someone who reads the business and builds the data?

Open to remote Finance Data Analyst, Analytics Engineer and FP&A automation roles. I answer in English and Spanish.

Disclosures

Built in public
This site and the projects behind it are documented as they are made, failures included. The engineering log records 30 defects found and fixed, numbered one by one.
Past results
The trading research shown here is a methodology demonstration, not investment advice.
Verifiable figures
Every number on this page comes from the pipeline or the public repository, and links to the artifact that proves it.
Power BI report
The report exists as a PBIP project in the public repository and opens for free in Power BI Desktop. There is no public embed because «Publish to web» needs a Pro licence on a work tenant and makes the dataset public.
Analytics
The site counts page views with Vercel Analytics: no cookies, no browser fingerprint and no personal data, served from this same domain. That is why there is no consent banner to dismiss.
Languages
Native Spanish · English B2 · Portuguese A2.