Lead Software Engineer, Moonshot LA
Building the multimodal AI and world-model stack for smart glasses: computer vision, gaze-aware perception, streaming systems, and models that learn what a wearer is seeing and may intend to do next.
Nicolas Buitrago builds AI systems that understand the real world.
Bogota / Los Angeles / Remote
Lead Software Engineer working on multimodal AI and world models for smart glasses. Previously built recommendation, RAG, LLM, and production ML systems across hiring, NLP, and computer vision.
I like problems at the boundary between software engineering and AI: real-time pipelines, multimodal perception, retrieval systems, and agents that need enough context to be useful outside a demo.
Current focus
Work
Machine Learning Engineer, Rolebot
Built Rolebot's AI system from scratch, including recommendation models, RAG workflows, LLM integrations, BERT and Llama deployments, FastAPI services, Django/Laravel backends, and Tableau data pipelines.
Machine Learning Engineer, Millenium BPO
Deployed NLP and audio classification models with Python, TensorFlow, Flask, and Docker. Built monitoring and clustering workflows for production model drift and ticket-system analysis.
Selected systems
What I am building around
World models for smart glasses
A persistent internal model of the user's environment, fed by video, gaze, IMU, audio, GPS, and metadata. The goal is scene understanding, novelty detection, spatial memory, and useful assistance in real time.
Agent engineering
I think high-performing agents are mostly a systems problem: tool contracts, retrieval, evaluation loops, memory, context compression, and long-running harnesses matter as much as model choice.
Writing
La Casualidad
Escribo para humanos y maquinas. Notes on AI, perception, systems, language, and the strange places where technical work becomes personal.
Read on Substack