Applied AI Researcher & ML Engineer
I'm Javier Montalvo, a Computer Vision PhD and Machine Learning Engineer based in Madrid. I build computer vision models from initial research to final deployment. My work translates research ideas into measurable systems through custom training frameworks, evaluation pipelines, and model optimization for deployment constraints.
Applied ML Engineering Impact
Production Computer Vision
At Identy.io, I build biometric security systems for both edge and server environments. I owned the full development of a Deepfake Detection module, and developed PAD models that achieved a 20x reduction in FAR, directly contributing to iBeta L2 certification and $5M+ in commercial contracts.
Model Efficiency
I engineer compact backbones distilled from foundation models to serve as the base architecture for downstream tasks. By targeting strict mobile hardware constraints, I enabled a 3x inference speedup and up to a 50% footprint reduction at the edge.
ML Systems Ownership
I build the training, evaluation, and deployment code around the model, not just the model itself. My responsibilities span the entire stack: sourcing data, designing architectures, and implementing production code. Recently, I consolidated legacy codebases into a unified training framework, cutting technical debt and accelerating the team's experimentation cycle.
Synthetic Data and Simulation
When real-world data is constrained, I build custom generation pipelines. I have used Unity, CARLA, and Stable Diffusion to generate synthetic datasets for visual perception tasks, reducing pose estimation error by up to 65% and boosting domain adaptation performance by 20%.
Technical Stack
Programming
Python, C++, C#, JavaScript
AI & Vision
PyTorch, TensorFlow, ONNX, OpenCV, Hugging Face, Transformers, Unity, CARLA
DevOps/Web
Git, Docker, AWS, Node.js, Next.js, MongoDB