ML Research Engineer

Mosaico is an open-source data platform for robotics and physical AI. We’re solving a hard infrastructure problem: the tools that exist today for storing, indexing, and retrieving high-frequency sensor data are either too primitive or too painful to work with at scale.

We’re building the missing piece, and doing it in the open. Apache 2.0 licensed, public roadmap, real community. We care about the codebase being something people actually want to read and contribute to, not just use. In Italy, companies that operate this way are still a rarity, and we think that’s exactly what makes this interesting.

We’re based in Reggio Emilia, working with teams in defence, automotive, and agritech, and we’re building a commercial offering on top of the open-source core.

The role

Mosaico is building semantic search over high-frequency, multi-modal sensor data (video, point clouds, time-series) and this role sits at the heart of that effort. You will work directly under the Senior Research Engineer for Semantic Search, turning architectural decisions into running experiments, trained models, and evaluated results.

This is not a pure implementation role. We expect you to understand what you are building and why, to flag problems early, and to contribute ideas. At the same time, the core of the job is execution: keeping the research pipeline moving, managing experiments rigorously, and delivering results the team can build on. If you are 2–4 years into your career and want to do real research engineering on hard problems in a small, technically serious team, this is the right place to grow.

What you’ll work on

• Implement neural network architectures for semantic search as designed by the Senior Research Engineer, from embedding models to retrieval pipelines
• Run training and fine-tuning experiments on LLM and Video-LLM models for visual and multi-modal semantic understanding
• Manage experiment tracking, evaluation pipelines, and systematic ablations to measure and improve model performance
• Contribute to the development of multi-modal embedding models across video, point cloud, and time-series data
• Collaborate with the storage and backend engineers on integrating trained models into the Mosaico platform
• Stay current with relevant research and bring relevant findings to the team’s attention

What we’re looking for

• Solid foundations in deep learning: you understand how models are trained, what can go wrong, and how to debug it
• Hands-on experience with representation learning or embeddings: training models whose output is a vector used for retrieval or similarity
• Comfortable implementing research papers and translating them into working, evaluated code
• Experience managing training pipelines and experiments in a rigorous, reproducible way
• Proficiency in Python and at least one major deep learning framework (PyTorch preferred)
• Able to work autonomously on well-defined tasks and communicate clearly when something is blocked or unclear
• Fluent in English, written and spoken

Nice to have

• Experience with multi-modal or cross-modal models
• Familiarity with video understanding or point cloud processing
• Experience with distributed training or large-scale training infrastructure
• Familiarity with vector search libraries and approximate nearest neighbor methods
• Open-source contributions in the ML or data space

What we offer

• Direct mentorship from a senior research engineer on hard, frontier problems in Physical AI
• A small core team working on hard infrastructure problems in the open
• Work that ships as open source under Apache 2.0 and powers a commercial product
• Flexible hours, full remote with optional office access in Reggio Emilia
• Competitive salary, discussed openly based on your level and experience
• Welfare package
• Stock options

Salary range

€35K – 50K, adjusted based on location, experience, and level.

Contacts

Interested? Write to [email protected]

Apply for this role

All positions