➜    rafaelrojasvi.github.io  git:( main )  ✗

Valdivia, Chile · open to remote / hybrid

Rafael Rojas

they/them · Junior Backend Software Engineer

Python · Elixir · APIs · Automation · Data Tools

“Brilliant and flawed, combining rare engineering talents with ridiculous public displays of ostentation, Musk is our era’s Thomas Edison – the man who, you may recall, electrocuted an elephant in order to discredit a rival.”

― Yanis Varoufakis, Technofeudalism: What Killed Capitalism

At a glance

  • Computer Science graduate with First Class Honours (University of Westminster, 2025). Studied abroad in London, United Kingdom.
  • Focused on backend development, automation, APIs, and data-driven tools.
  • Strong with relational databases, integrations, Python applications, API integrations, and automation workflows.
  • Currently: Software Engineer (Data & Automation) at OrigenLab; freelance backend and integration work (London / remote).
Projects & case studies Experience Email

Selected work

Production backends & integrations

Four builds with different constraints: employed data/automation, freelance payments + messaging, academic ledger API, and an NLP evaluation project. Each write-up calls out stack, ownership, and where reliability mattered.

Project

OrigenLab

Production full-stack operations platform: Astro public site, Python email intelligence pipeline, SQLite operational store, read-only FastAPI, optional Postgres mirror reporting, and React operator dashboard

Production (employed)

Four-app monorepo: origenlab.cl (Astro), Python ingest/marts (SQLite truth), read-only operator API (:8001), and React operator dashboard—with optional Postgres mirror reporting and human-reviewed outreach gates.

Highlights

  • Public site: product-forward homepage, Ortoalresa centrifuge vitrine + comparison, SERVA reagents, catalog validator
  • Brand: three-body logo (canvas header), static lockups, SVG tooling, conservative commercial copy governance
  • Pipeline: SQLite authoritative store; optional Alembic Postgres mirror for reporting; email-pipeline owns all writes
  • API & UI: GET-only apps/api (operator routes + /mirror/* reporting); multi-section React operator dashboard + smoke tests
  • Lead intelligence: SQLite lead research imports, Postgres lead_intel mirror, GET /mirror/leads/*, read-only Prospectos page
  • Safe outbound: sent-history, suppression, duplicates, human-reviewed LLM drafts—no auto-send

Stack

Astro 5 Tailwind CSS 4 TypeScript Python SQLite PostgreSQL Alembic +5

Project

Tattoo Booking Bot

Freelance platform: FastAPI, PostgreSQL, Stripe, and Meta WhatsApp with production-style webhook discipline

Production (freelance)

End-to-end booking stack: WhatsApp webhooks into FastAPI, durable conversation/payment state in PostgreSQL, Stripe deposits with webhook reconciliation, and Google Calendar/Sheets integrations—with pytest on critical paths and Docker + GHA for delivery.

Highlights

  • Idempotent WhatsApp webhooks & duplicate-safe Stripe reconciliation
  • PostgreSQL, SQLAlchemy, Alembic migrations; durable session + payment state
  • 13-question consultation flow with validation; admin / artist approval before commit
  • Docker, pytest, GitHub Actions; production safeguards on critical paths

Stack

FastAPI PostgreSQL SQLAlchemy Alembic Stripe WhatsApp Business Cloud API Google Sheets +4

Project

LedgerBank API

University-grade Phoenix API: JWT auth, PostgreSQL ledger, Oban jobs, Docker/CI

Academic / portfolio

Banking-style Elixir/Phoenix API: JWT authentication, transactional ledger modelling in PostgreSQL, Oban background jobs, and clean architecture—focused on invariants and backend reliability, not UI polish.

Highlights

  • JWT-secured routes with role-aware permissions
  • Transactional account/transfer modelling
  • Oban for async, retried jobs
  • Docker + CI for reproducible builds

Stack

Elixir Phoenix PostgreSQL Oban JWT Docker CI/CD

Project

Political Misinformation Classifier

Applied ML: BERT fine-tuning, evaluation discipline, honest limitations

Academic

Python NLP project: BERT-based political misinformation / fake-news classification with data preprocessing, train/validation discipline, model evaluation, and explicit limitations for responsible use.

Highlights

  • Tokenizer-aligned preprocessing
  • Train/val rigor and confusion analysis
  • Reproducible scripts & data hygiene
  • Responsible framing (limitations first)

Stack

Python BERT NLP Machine learning Model evaluation Data pipelines

All projects →

How I work

What I’m strongest at on a backend team

Services & data models

HTTP APIs and schemas with explicit validation—so invalid state fails in one place, not across three integrations.

Automation with guardrails

Pipelines and outbound workflows with suppression, history checks, and human review where risk is high.

Integration-heavy backends

Stripe, Meta, Google: webhooks, retries, timeouts, and tests around the paths that move money or messages.

Operator-facing tools

Exports, dashboards, and runbooks so the team can answer “what happened?” without opening raw logs. 

# What I default to on backend work
engineering_default = %{
  title: "Engineering default",
  note:
    "I optimize for backends another engineer can extend: explicit data rules, migrations you can trust, and tests on parsers, webhooks, and payment edges—where production usually breaks."
}

iex(1)>   Portfolio.EngineeringDefault.run()