Backend & distributed systems engineer based in Pasto, Colombia.
I build async, event-driven services for scientific data. Right now that's Lakiy, a volcanic & seismic monitoring platform at the Servicio Geológico Colombiano (SGC), Colombia's national geological survey — a system that has to stay correct and responsive under real-world load, where the data doesn't forgive mistakes.
I like systems thinking at every level — from event-driven architecture down to writing a 3D graphics pipeline from scratch just to understand it properly.
- Backend — Python (Django · DRF · async/ASGI · Celery), FastAPI, Go, Java/Kotlin (Spring Boot)
- Data & streaming — PostgreSQL, MongoDB, Redis, ScyllaDB · Kafka · gRPC
- ML serving — TensorFlow Serving, MLflow, MinIO/S3
- Frontend — React + TypeScript (Vite), real-time Canvas rendering
- Infra — Docker, GitLab CI/CD, NGINX
🌋 Lakiy · closed source (SGC) — the volcanic & seismic monitoring platform above, which I build full-stack. Polyglot microservices (Python · TypeScript · Kotlin) behind an NGINX gateway, Dockerized and shipped through GitLab CI/CD.
- Backend — async-first Django 5 on ASGI/Daphne (DRF, Channels/WebSockets, Celery), event-driven over Kafka (aiokafka + Apicurio Schema Registry); multi-database PostgreSQL with a routing layer bridging modern and legacy
managed=Falseschemas; TCP integration with legacy Java detection/localization servers and seismic data sources (Earthworm, FDSN, SeedLink, Winston). - Frontend — React 18 + TypeScript SPA with real-time Canvas rendering of seismograms, spectrograms, helicorders and location maps (no charting library), Recoil/Jotai state over WebSockets.
- Scientific services — waveform image rendering in Spring Boot (Kotlin) with multi-level Caffeine + Redis caching and source fallback, plus a FastAPI + Matplotlib service over MongoDB; signal processing with ObsPy and neural-network classification of volcanic events.
- LLM tooling — an MCP server (FastMCP) exposing seismic queries and geospatial maps.
Closed source, but I'm happy to walk through the architecture.
🔻 kenderer — a 3D rendering engine in pure Go with two interchangeable backends (a from-scratch CPU software rasterizer and an SDL_GPU / Vulkan path), kept in lockstep by differential parity tests. A deliberate deep dive into graphics, linear algebra, and low-level performance.
🌐 schema_scraper · goceptron — smaller Go projects (web scraping, a from-scratch perceptron).
Systems engineer (Ingeniero de Sistemas, Universidad de Nariño). Away from the editor: specialty coffee (V60 pour-over), sim racing, and anime.