Sebastian Velandia Acevedo

I build software that turns manual work into reliable systems: AI agents and LLM pipelines that handle repetitive tasks end-to-end. I designed and built a generation pipeline for actas, separating what an LLM decides from what is derived deterministically—so the output doesn’t hallucinate administrative data.

I’ve applied the same logic in real environments, including work for a regional hospital, and I built an agentic recruiting search platform for production use in 48 hours through Kiami (GTM Hackathon with Anthropic and Vercel). I also develop complete products in Next.js for real clients.

Right now, what excites me most is designing and building AI agents and orchestration systems—where reliability and clear separation of “model decisions” vs “deterministic derivation” are first-class. Outside the code, I study Electronic Engineering and run a hands-on experimental rocket project (data-acquisition firmware, flight simulation) with an active contribution to OpenRocket.