Pento is a team of Machine Learning and Software Engineers who design, build, and deploy AI systems that solve real business problems. From semantic search and recommendation engines to computer vision and generative AI, we deliver end-to-end — from research and prototyping through production.
Our engineering team brings deep hands-on expertise across ML, full-stack development, and cloud infrastructure. We integrate directly into client teams and operate in North American time zones.
We contribute tools and research back to the community, with a focus on AI-powered search, ML workflows and AI-native engineering:
- quality-gates-skills — An agent skill that sets up pre-commit hooks and CI checks for TypeScript and Python repos: format, types, tests and coverage, complexity, duplication and dead code. Your coding agent calibrates them to the code you have and proves they block a bad commit. Works with Claude Code, Codex, Cursor and skills.sh.
- minicoil-en-es — miniCOIL EN-ES, co-built with Qdrant: a sparse retriever that matches English and Spanish in a single index, with no translation step. Model on Hugging Face.
- alphaXiv MCP — With alphaXiv: an MCP server that brings alphaXiv's research corpus into Claude Code, Cursor, and other AI coding tools. See from research paper to production.
- ml-ralph — An autonomous ML research agent that runs experiments, maintains a living plan, and learns as it goes.
- vectory — Track and compare embedding versions. Essential tooling for teams iterating on search and retrieval systems.
- streamlit-terran-timeline — A Streamlit component that builds face-recognition timelines for videos.
