Looking for a job

alexander korabelnikov User

@alexanderkorabelniko

AI inference optimization researcher at Mila4AI who cut inference and infrastructure costs 2–3× for a speech-therapy AI platform.

Cyprus

Message

At Mila4AI, I optimized inference pipelines for a speech-therapy AI platform, reducing inference and infrastructure costs by 2–3× through model compression, hardware selection, and client-side inference. I also accelerated critical ML pipelines by 4× using neural network optimization and TensorRT and Triton deployment.

At Docet TI, I led two teams of developers and designed perception pipelines for autonomous vehicles, including 100fps sensor fusion and 50fps camera neural networks on edge accelerators. At Arrival, I designed, trained, optimized, and deployed CNNs for ADAS functions on embedded hardware with strict latency and memory constraints.

My work spans inference optimization for large-scale neural networks, from model compression and hardware-aware acceleration to production deployment. I have worked with transformer architectures, computer vision, and real-time systems, and have a PhD program background in computer science at HSE.

Experience

Work history, roles, and key accomplishments

Current

Head of Software (ML/AI)

Current

Mila4AI

Jan 2023 - Present (3 years 9 months)

Optimized inference pipelines for a speech-therapy AI platform, reducing inference and infrastructure cost by 2–3× through model compression, hardware selection, and client-side inference. Accelerated critical ML pipelines by 4× via neural network optimization and achieved 50–150% speedup in production systems.

Model Compression TensorRT Triton Inference Optimization Neural Network Optimisation GPU Optimization Client Side Inference Production Systems

Lead ML Engineer (Autonomy & Perception)

Docet TI

Jan 2022 - Dec 2023 (1 year 11 months)

Led two teams of middle/senior developers in computer vision and reinforcement learning, choosing tech approaches and software/hardware components. Designed and optimized perception pipelines for autonomous vehicles, achieving 100fps sensor fusion and 50fps camera neural networks on edge accelerators.

Computer Vision Reinforcement Learning Sensor Fusion AI Hardware Accelerators PCL Octree Compression H.264 H.265 Self Supervised 3D Detection Cross Domain Distillation

Lead Computer Vision Engineer (ADAS & Edge AI)

Arrival

Jan 2019 - Dec 2022 (3 years 11 months)

Designed, trained, optimized, and deployed CNNs on embedded edge hardware meeting strict latency and memory constraints for ADAS functions in autonomous vehicles. Managed full-cycle inference optimization including model compression, hardware selection, deployment, and benchmarking.

PyTorch CVAT C++MMDetection TensorRT TIDL Model Compression Edge AI ADASエンジニア Sensor Fusion LiDAR Camera Calibration

Teaching Assistant

HSE

Jan 2018 - Dec 2019 (1 year 11 months)

Assisted in teaching a class about ConvNets during PhD enrollment.

Convolutional Neural Networks Teaching

Lead Researcher

Aligned Research Group LLC

Jan 2014 - Dec 2019 (5 years 11 months)

Conducted medical image processing and segmentation on CT scans using classical ML, CNNs, and 3D shape models. Developed face recognition and person re-identification systems using CNNs, metric learning, and ensembles.

Medical Imaging CT Imaging Segmentation Traditional ML CNN 3D Shape Models Face Recognition Person Re Identification Metric Learning C++CUDA OpenMP Caffe

ML&CV Tech Leader

NDA

Jan 2018 - Dec 2018 (11 months)

Launched an R&D project for a surveillance system, including requirements, estimates, and business goals. Developed object detection and tracking prototypes, hired developers, supervised, and transferred the project.

Object Detection Object Tracking Prototyping Team Management R&D