Jose Esparza
I engineer large language model systems and the infrastructure that keeps them running in production: agentic architectures, retrieval-augmented generation, supervised fine-tuning and low-bit quantization. I am a Senior Artificial Intelligence Engineer at Subatomic and hold a concurrent contract at Maximo. I also founded Hoja, an AI and software consultancy.
For seven years before that I built machine learning and data systems across education, fintech, retail, mining and the public sector, first as an engineer and later leading teams. You will find those years on my resume, and the systems and publications on the portfolio page.
I am completing an MSc in Computer Science at the University of London (expected December 2026), and I graduated in Software Engineering from Universidad Privada del Norte with a 4.0 GPA. I started research at Binghamton University in Arti Ramesh's lab, where Gisella Bejarano mentored me; we published our work on public discourse around COVID-19 vaccination at CSoNet 2021.
In Julia I maintain open-source scientific software: MLJ.jl, DearDiary.jl under the JuliaAI organization, and MichiBoost.jl. Google Summer of Code funded my work bringing MLflow to the ecosystem in 2023.
Outside work I sing in El mejor verano de mi vida, play classical guitar and Magic: The Gathering, and keep two cats.
Research interests
- Language models
- Agentic architectures, retrieval-augmented generation, supervised fine-tuning
- Efficient inference
- Low-bit quantization (NVFP4 / W4A4), on-device and CPU-only deployment
- Representation learning
- Topic modeling, semantic retrieval, recommender systems
- Scientific computing
- Open-source machine learning tooling for the Julia ecosystem
Technical stack
- ML & LLMs
- PyTorch, HF Transformers, vLLM, llm-compressor, LLaMA-Factory, LangGraph, LlamaIndex, MCP/FastMCP
- Data & infra
- Spark, Airflow, Databricks, AWS, GCP, PostgreSQL/pgvector, Docker, Kubernetes
- Languages
- Python, SQL, Julia, TypeScript, C++, CUDA