Anton Alekseev
MLOps engineer. I build ML and inference platforms on Kubernetes and make sure GPUs never sit idle.
Where I have worked
Now · since 2025
Large classifieds marketplace
I develop the ML platform on Kubeflow: I work on raising GPU utilization and brought in Kueue for batch jobs and Vault for secrets. In parallel I build the inference platform on KServe: Knative and Standard modes for ML and LLM models, plus a service for distributed disaggregated LLM inference.
2022-2025
Cloud provider
Built an MLOps platform on Kubernetes with ClearML, JupyterHub, Keycloak and Istio, and a marketplace of ready-made Data Science images with CUDA. Developed an inference platform on Triton with autoscaling. Researched how GPU sharing (MIG, time-slicing, MPS, HAMi) lowers the cost of GPU flavors.
Projects
Bank
Concept of an ML platform on ClearML, Feast, JupyterHub and KServe.
Course and channel
Author of the course "MLOps for model development and monitoring" at Yandex Practicum. Mentor on getMentor. I run the Telegram channel "AI & ML Infrastructure" and write on Habr.