Juan Garassino — AI Engineer
Juan Garassino is an AI engineer based in Berlin, Germany, specializing in
multi-agent systems, large language model (LLM) applications, and generative AI. He designs
and ships production AI systems and teaches machine learning, with a background that bridges
architecture (Universidad de Buenos Aires) and applied AI/ML engineering.
What Juan does
Juan builds multi-agent orchestration systems (LangGraph, LlamaIndex, and custom hand-rolled
agent loops), retrieval-augmented generation (RAG) pipelines, fine-tuned and quantized LLMs
(LoRA/QLoRA, PEFT, RLHF with GRPO/PPO/DPO), and generative models (diffusion, GANs). He works
across the full stack — FastAPI backends, Docker/Kubernetes, and GCP, AWS, and Azure — and has
a research interest in LLM interpretability and AI red-team security.
Current roles
- Forward Deployed AI/ML Engineering Consultant — Engenious Inc, Berlin, Remote (US) (2026–present)
- Data Science & ML Engineering Coach — neuefische, Berlin (2026–present)
- Founder & Generative AI Engineer (freelance) — artSquared Studio, Berlin (2021–present)
- Lead Data Science & AI Instructor — Le Wagon, Berlin, taught in Berlin, Tokyo, Amsterdam, Kuala Lumpur, London, Abu Dhabi, Munich, Barcelona, Riyadh (2021–present)
Core skills
Python, PyTorch, TensorFlow, Transformers, LangGraph, LangChain, LlamaIndex, Hugging Face,
FastAPI, multi-agent systems, RAG, LLM fine-tuning, MLOps (MLflow, Weights & Biases),
vector databases (ChromaDB, Pinecone, FAISS, pgvector), and cloud ML (Vertex AI, SageMaker).
Work with Juan
- Engineering — short-term missions, MVP builds, and technical advisory.
- Recruitment — full-time roles, talent acquisition, and candidate analysis.
The site's Prints section shows hand-curated pen-plotter plates generated with PromptPlot,
Juan's LLM-to-GCode drawing pipeline for pen plotters.
Links
