Experteer Overview In this role you will own and advance models that read handwritten student input on digital boards, enabling real-time understanding and feedback in the classroom. You will shape architecture, input representations, and training methods to handle variable numbers of handwritten objects. You’ll work with a small team close to product decision-makers to ensure robust, deployable solutions that perform in school environments. Your work directly influences how students learn math with AI-assisted handwriting interpretation, delivering measurable classroom impact.Compensaciones / Incentivos
- Own the models: architecture, input representation, instance proposal/assignment, training, and iterative improvement at scale
- Design and implement changes to address hard cases and ensure trustworthy results
- Maintain a fair experimental framework: predefined success criteria, clear trade-offs, and rigorous reporting
- Deliver production-ready models: fast on standard school hardware,



with reliable deployment and rollback mechanisms
- Collaborate with datasets/evaluation owner to target impactful improvements
- Translate research findings into deployable improvements that survive real-world student data
- Document experiments and ensure reproducibility, including failed attemptsResponsabilidades
- Real depth in multiple-instance detection or instance segmentation
- Experience with non-rasterized input (strokes, polylines, trajectories, etc.)
- Attention-based architectures over sets and graphs; understanding spatial/relational encoding
- Serious experience training deep models in PyTorch (not just using APIs)
- Experimental discipline for fair comparisons and noise control
- Ability to take models into production: export, latency, determinism, rollbackRequisitos principales
-

📌 Senior Computer Vision Engineer (Barcelona)
🏢 Innovamat
📍 Barcelona