Ml Software Developer
CurrentResponsible for developing Grazper’s model training pipeline, training and optimizing pose estimation and box detection models central to our human pose recognition system.Developed and maintained a 3D human-pose annotation web application, now actively used by annotators to label our in-house dataset.Built an auto-annotation system using state-of-the-art models and postprocessing to accelerate the annotation process.Automated key steps in the recording pipeline (actor registration, video export, and auto-annotation) through custom scripts and applications.Led the migration to GitHub, establishing reproducible test environments for Grazper’s CI/CD pipelines on self-hosted GitHub runners. Deployed a self-hosted package registry for managing Python and npm packages, and Docker images, improving significantly the development process of the entire company.
