Vladimir Poliakov
About
Detail
Head of Engineering (Vision AI)
Bavaria, Germany
Résumé
Jobs
verified_user
0% verified
Head of Engineering (Vision AI)Sword HealthJan 2026 - Current (8 months)- Staff ML EngineerKaia HealthOct 2021 - Jan 2026 (4 years 4 months)At Kaia Health, I've built AI-powered technologies that drive innovation in personalized digital health. My work spanned the development of advanced 3D computer vision systems, applying large language models (LLMs), and driving product-focused solutions. I provided technical leadership in integrating AI and machine learning into healthcare products, building diverse ML solutions that enhance user engagement and improve personalized care delivery.
- ML Engineer / Tech LeadKaia HealthApr 2020 - Sep 2021 (1 year 6 months)
- ML EngineerKaia HealthApr 2018 - Mar 2020 (2 years)
- ML Engineeriteratec GmbHNov 2016 - Apr 2018 (1 year 6 months)• Expanded the company’s AI portfolio by delivering 5 projects on diverse topics, including text anomaly detection, sentiment analysis, on-device object detection, pattern recognition, and semantic segmentation. • Developed an internal service to identify and analyze trending topics from millions of news headlines sourced from popular social media platforms. Mentored a team of two working students.
Education
verified_user
0% verified
- Master of Science - MS, Computational Science and EngineeringTechnical University of MunichJan 2016 - Jan 2018 (2 years 1 month)Specialization in Computational Statistics and Mathematical Biology.
- Bachelor of Science - BS, Applied MathematicsBauman Moscow State Technical UniversityJan 2011 - Jan 2015 (4 years 1 month)Specialization in Mathematical Simulation.
Projects (professional or personal)
verified_user
0% verified
- IInferring 3D Human Pose in Real-Time on Consumer SmartphonesSep 2018 - Feb 2019 (6 months)Master's Thesis. In this thesis, two lightweight three-dimensional human pose estimation models are presented. These models are based on the idea of the stacked hourglass network, a deep convolutional architecture, optimized for the mobile devices. The first model transfers two-dimensional points to three-dimensional space using the shallow neural network with residual structure. The second weakly supervised approach combines an optimized two-dimensional stacked hourglass model with a depth regression module in a single deep neural network. As part of the thesis, various configurations of hourglass architecture blocks based on convolutional layers were analyzed and tested. It has been established that the developed lightweight models are c
- UU-net image segmentation Kaggle competitionJul 2017 - Sep 2017 (3 months)In this project, a system that automatically removes cars from the photo studio background is developed. The project is designed as a part of Kaggle competition for online used car startup called Carvana. Such a model will allow Carvana to superimpose cars on a variety of backgrounds. The result of a competition is an ensembled model of different approaches made in a group of 3 persons. The Public Leaderboard place: 10-th. The Private Leaderboard place: 51-th (top 7%). The Private score was lowered due to the presence of exactly one car in the Private set on which the model crashed. In all other respects, the overall model is very efficient. The mistake made will be a good experience for us.
- AA dynamic hierarchical Poisson model for the prediction of soccer match outcomesMay 2017 - Aug 2017 (4 months)In this term project, a dynamic hierarchical Bayesian model is proposed for analyzing and predicting the results of a football match, which is assumed to come from a Poisson distribution with teams abilities that change stochastically over time. The No-U-Turn sampler algorithm is used to solve the inverse Bayesian problem of finding such abilities, i.e. attack and defence coefficients distributions. Example validation of the model is performed on the Italian Serie A championship 2016-2017. The performance of the model is also tested on a naive betting strategy that applies to the results of the 10 top league matches in the 2016-2017 season.