Eugenio "Jay" Zuccarelli

agentic ai // generative models // production ai
retrieval-augmented generation // evals // rlhf

supervised // unsupervised // reinforcement learning
feature engineering // model selection // hyperparameter tuning

experimentation // causal inference
forecasting // decision systems

open harness for proactive ambient agents that watch, listen, and only speak up when it's worth your attention. local model runs continuously on-device for privacy; remote model only fires for harder calls.

an AI gardener. a Claude agent tends a real hydroponic basil bucket through Home Assistant, with a deterministic reflex tier underneath so the plant survives the model's mistakes.

AI-powered form autofill agent built on browser-use — describe yourself once, point it at any web form, and it fills every field automatically. you review and submit manually.

persistent memory for Home Assistant Assist, in plain markdown. no vector DB, no embeddings — works with Anthropic Claude, OpenAI, and any LLM conversation agent.

personal cross-LLM memory layer over MCP. markdown files are the source of truth; Claude, ChatGPT, and Cursor share one memory.

Eugenio "Jay" Zuccarelli is an AI-native operator who builds agentic AI systems and ships machine learning models in production — across the full lifecycle from research to deployment.

agentic AI and applied ML — autonomous agents, generative models, evals, and decision systems. open-source projects include saccade (ambient agent harness), autofill (form-filling agent), and ha-memory (memory for Home Assistant).

python, pytorch, and tensorflow on aws and gcp, with sql, spark, docker, and kubernetes.

github, x, and linkedin — all @jayzuccarelli.