“Engineering is the art of understanding complex systems.”
I'm Sebastián López Osorno, a developer from Medellín, Colombia, in my final semester of Computer Engineering. For more than 4 years I've been getting systems that don't talk to each other to share data people can trust.
My infrastructure work started at MUV-U, with the company's servers and support for its users. At ADCOM I set up a VPS with Docker, Nginx and SSL, and later the deployment of the landing pages, with a user without a shell that uploads the files through rsync without restarting anything.
At TIVENOS I publish 20+ services to GHCR. Since I needed to hear about a failure before a team reported it, I built the stack that watches them, with metrics in Prometheus, logs in Loki, traces in Tempo and Grafana alerts that reach Telegram.
Along that path I've worked on almost every side of a tech team, from data, CRM and backend to frontend, servers and, at TIVENOS, technical coordination. It lets me understand what marketing, finance or IT ask for without anyone translating it for me, and turn it into concrete tasks.
With artificial intelligence I took the same path, from building it to directing it. At university I built, with a classmate, an expert system in CLIPS, fuzzy logic with scikit-fuzzy and genetic algorithms, and on my own a reactive agent that decides which actuators to switch on; at SmartPot, an assistant with expert rules, fuzzy logic and scikit-learn models that forecasts when a crop will leave its range and proposes actions applied in bulk; and at ADCOM, lead contact with AI agents in Dapta.
I work with language models at both scales. On the small side, I fine-tuned Qwen2.5-1.5B with LoRA and built a RAG pipeline with a quantized Qwen2-0.5B and a FAISS index, meant to run on modest hardware. On the large side, I built an n8n agent that answers with Claude and queries three MCP servers I wrote in Python, and I've worked with GPT, with Gemini and with open models running locally through Ollama.
In daily development I direct coding agents, Claude Code among them, with engineering judgment. Before delegating I write down each project's rules, from commit authorship to what is never touched in production; then I review every proposal and run it through the same tests and the same CI as any other change. Having built these systems from the inside is what lets me decide what to delegate, spot when an answer doesn't hold up and use AI to save work without giving up the design.
Next is bringing those language models to the data I integrate today, with assistants that answer from each organization's own information. On the academic side, I finish the cybersecurity diploma in December 2026, graduate in March 2027, and then come a specialization in data science and a master's in cybersecurity.
What I care about most is that the teams depending on a service are never the first to find out it went down. I'm looking for a DevOps or platform role, remote or hybrid.
