Damir Murtazin
Senior Ml Engineer Data Scientist @Viridien
Signup · Get unlimited contacts
WORK HISTORY
Senior Ml Engineer Data Scientist @Viridien
Crawley, GB
Delivered an end-to-end instance segmentation model for ultra-large images (>2 billion pixels) within 5 months, achieving an IoU score of 0.89. Key contributions included designing an automated labeling pipeline, selecting and training the optimal model, and optimizing multi-GPU inference workflows- Fine-tuned the multimodal VLM LLava (image + text + tabular data) to generate detailed core sample descriptions from electron microscope images, cutting human processing time from 3 hours to just 30 seconds while maintaining accuracy- For one of the computer vision project: decreased RAM consumption (750 GB -> 50GB), decreased inference time by 2x and adapted a zero-shot model for highly specific industrial task that allows to save 4 months on dataset labeling and HPC resources for model training. Results is up to 400 unique segments for one image with high accuracy. With this projectI have successfully secured a contract representing 20% of last year’s volume- Engineered distributed ML pipelines leveraging DeepSpeed and PyTorch DDP to utilize 80+ GPUs across 10 nodes for both training and inference in two major projects- Host regular knowledge sharing sessions for colleagues on the following topics: heavy computation optimization on python, DS projects reproducibility, DS for entry level- After one year in the company won \"Innovation award\" for one of the project above- Onboarded 3 new hires.
EDUCATION
Perm State University (PSU)
Master's degree, Geophysical study of the Earth's crust
Perm State University (PSU)
Bachelor's degree, Prospecting of Mineral Resources
Ufa State Petroleum Technical University
Doctor of Philosophy - PhD, Engineering
SKILLS
ABOUT DAMIR MURTAZIN
Experienced Machine Learning Engineer with 4+ years of expertise in delivering impactful AI solutions for large-scale industrial challenges. Specializing in computer vision and distributed systems, I excel at building scalable ML pipelines and optimizing performance for high-complexity tasks.Key achievements:Developed an instance segmentation model for ultra-large images (>2 billion pixels), achieving an IoU score of 0.89 while optimizing multi-GPU inference.Reduced RAM usage by 15x and halved inference time in a computer vision pipeline, enabling up to high-accuracy segments per image. Engineered distributed ML pipelines across 80+ GPUs on 10 nodes using DeepSpeed and PyTorch DDP for scalable training and inference.Passionate about innovation and knowledge sharing, I actively mentor colleagues and onboard new team members to drive success in cutting-edge AI projects.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.