Pavel Solomein

- Computer vision system to detect and track objects at the steel factory shop floor (trains, buckets, shovels, cranes). Improved productivity and coordination between shop floors. [PyTorch OpenMMLab / YOLOX, docker] - Prototype to identify products on a supermarket shelf from images. Best accuracy achieved ~97% [PyTorch YOLOv5] - Model to detect defects of a superconductor tape with a linear camera. Random quality checks replaced with 100% automated batch inspection [TensorFlow U-NET, flask, docker]

- Co-founded FoodTech startup aiming to improve ordering experience at restaurants - Created telegram bot prototype for voice ordering at a restaurant using ML [spacy fuzzy search, google speech-to-text, django] - Established relationships with the pilot restaurant, investors and partners - Co-managed a team of 3 freelancers to release the mobile app (“Кушац” in App Store / Google Play)

- Consulted customers and built ML models for industries: banking, retail, manufacturing, chemical, metallurgy. Specialised in predictive quality/maintenance problems. [SAP HANA, R, SQL, pandas, sklearn, xgboost, flask, docker] - Won internal competition "SAP Innovator Challenge 2018". Assembled and managed a team of 6 colleagues to build an innovative IoT/ML/Cloud prototype solution for CrossFit [TI SensorTag, raspberry pi, MQTT, sklearn, flask, paper implementation]

- Modified source code of a packaging machine to analyse breakdown data, reducing stoppages by 20% - Managed a transition to breakdown root cause analysis system in factory, reducing breakdown time by 2%