Deep Learning for Vision
At Chute, I spent two years developing production machine learning systems that analyzed images shared online. I implemented the end-to-end machine learning pipeline and trained specialized deep learning models to recognize brands, objects, and scenes.
These systems helped companies identify relevant content and social marketing opportunities at scale. My work connected model development, data processing, and the infrastructure needed to run image understanding reliably in production.
I also planned and executed the migration of the data infrastructure to a fault-tolerant, distributed streaming architecture built on Apache Kafka. My daily tools included Python’s data ecosystem, Redis, Caffe, and NVIDIA DIGITS. As TensorFlow emerged, we also evaluated it as a possible framework for the team’s work.
The Chute Chat product was later acquired by another company.