Adding tools to curate datasets and algos to break models in months made me tech lead
Startup founded by the inventor of ELMs, specializing in low-power, edge-deployed AI models. I advanced from junior engineer to team lead by spearheading key improvements to our object detection pipeline. I developed in-house tools, including semi-supervised data labelling systems, hyperparameter search automation, and optimizer improvements. Through scrapped data integration, augmentation techniques, and algorithmic enhancements, I improved our best model’s mAP for most applications from 0.70 to 0.85. These efforts also laid the foundation for productizing our object detection solutions.