Fullstack engineer moving toward ML systems. I build data-heavy backends and the interfaces on top of them, and I like problems where the win is measurable: a 10‑minute geometry calculation brought down to 10 ms, a week-long aviation report turned into a one-day job, a sign-language dataset that didn't exist until I recorded it.
Based in La Paz, Bolivia 🇧🇴 · Software Developer at Helios Software · dougrojas.com
- Helios FHIR Server (open source) — built the MongoDB backend for the SQL‑on‑FHIR engine, wrote pysof (Python bindings for the Rust query engine), and fixed edge cases in the HL7 FHIRPath validation engine.
- Healthcare claims evaluation platform — fullstack work on a system that parses and validates X12 EDI transactions (837/381) across 30+ US states. Rebuilt the batch validation pipeline and reporting, and integrated an external claim‑validation API so secondary claims get verified automatically.
- Self‑study in ML — transformers, reinforcement learning, and optimization theory, with the goal of contributing to applied AI research.
| Project | What I did | Stack | Result |
|---|---|---|---|
| Shade analysis engine | Rewrote the projection and linear‑algebra core of a solar shade calculator running on Kubernetes | Python, NumPy, GKE | 10 min → 10 ms per run; large cut in compute cost |
| PDF report pipeline | Reworked business logic, resource allocation and headless rendering for large generated reports | Node, Puppeteer, GCP | Processing time halved (8 → 4 min) |
| Bolivian Sign Language Recognition | Recorded the first dynamic Bolivian Sign Language dataset and trained a gesture‑recognition model on it | TensorFlow, Keras, MediaPipe, OpenCV, PyQt | 87% accuracy on 9/10 gestures |
| World Magnetic Model Report Generator | Built a desktop tool that computes magnetic declination reports for airport procedure design | Python, Pandas, NumPy, Tkinter/PyQt | Evaluation time 1 week → 1 day, adopted by civil aviation |
| Sports social‑network MVP | Co‑led a 3‑dev team from architecture to launch in 3 months: Ethereum smart contracts, React front end, serverless video watermarking | React, Redux Toolkit, Solidity, AWS S3/Lambda, ffmpeg | Shipped on schedule |
| Backend DX & CI/CD | Set up GitLab pipelines with coverage gates and automated GCP deploys; cut build times; onboarded new devs | Node, GitLab CI, Docker, GCP | 70% coverage threshold enforced |
| CRA → Vite migration | Migrated a React app off Create React App, introduced Redux Toolkit, and refactored a backend‑shaped UI into a proper frontend architecture | React, Vite, Redux Toolkit | Faster builds, cleaner state |
Languages — TypeScript/JavaScript, Python, Rust
Backend — Node.js, NestJS, Express, FastAPI · Postgres, MongoDB, DynamoDB · TypeORM, Sequelize
Frontend — React, Angular, MUI, Cypress
ML / data — TensorFlow, Keras, OpenCV, MediaPipe, NumPy, Pandas, Open3D
Cloud / DevOps — Docker, Kubernetes, GCP, AWS, GitLab CI/CD, Prometheus
B.Sc. Computer Science & Mechatronics Engineering, Universidad Católica Boliviana (2019, graduated with distinction). Aspire Leaders Program 2025. Before software: a year in procurement across Bolivia, Paraguay and Chile, which is where I learned to write a report people actually read.
Off the keyboard: soccer, lifting, and philosophy.
