Evgenii Markin

I've built production AI systems at Celsus for medical imaging, taking stroke analysis from research through clinical deployment. My platform detects hemorrhagic and ischemic stroke and automates brain morphometry, processing thousands of CT studies each month across dozens of hospitals.

I design task-specific deep learning architectures, training strategies, and robust clinical-data pipelines using PyTorch, Triton Inference Server, ONNX, and Docker. I work directly with radiologists, support validation and regulatory documentation, and lead a team of four engineers while remaining hands-on with model development and architecture.

Alongside medical imaging, I've developed video analytics systems for person re-identification, object detection, action recognition, and age estimation.

At Alif, I'm developing banking AI solutions including a multilingual OCR platform for financial documents and identity-verification systems. I created a privacy-conscious synthetic data pipeline for Tajik OCR, achieved a 0.01 character error rate on real documents, and replaced a proprietary OCR solution to reduce costs by about $6,000 per month.