Description

Computer vision engineer with 8+ years of experience building imaging and ML pipelines for scientific and commercial applications. I design end-to-end systems — from dataset preparation and model training to API deployment — in Python and PyTorch. My academic background (MIPT MSc + PhD research) gives me a genuine edge on complex, domain-specific imaging problems where generic solutions fall short.

Open to short missions and longer consulting engagements: object detection, image segmentation, OCR pipelines, 3D volume processing, scientific image analysis, data pipelines. I write clean, well-documented code and deliver on time.

Languages

English
Native or bilingual
Russian
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

Institute of Fundamental Problems of Biology, Russian Academy of Sciences
Computer Vision Engineer
January 2023 - Today (3 years and 6 months)
– Built automated microalgae cell detection & counting pipeline (YOLOv8 fine-tuned on in house annotated dataset) — fully replaced a manual workflow that previously consumed hours per sample
– Developed 3D chloroplast reconstruction pipeline from confocal microscopy stacks using custom mesh processing in Python
– Established image preprocessing standards; annotated 800+ images for model training
– Mentored junior researchers on image analysis and Python tooling
Python Data science Computer Vision
NanoCar
Computer Vision Consultant FREELANCE
March 2026 - May 2026 (2 months)
– Developed a product shelf presence detection system: YOLO fine-tuning on a custom retail dataset (~2,400 annotated images, 14 product categories)
– Built a data augmentation pipeline (synthetic occlusion, brightness shifts, perspective) to compensate for limited training data — improved mAP by ~18%
– Delivered lightweight REST inference endpoint (FastAPI) with confidence thresholding and per-category reporting; integrated into client's backend
IsPL
ML Engineer — Document Processing FREELANCE
May 2026 - May 2026
– Built an OCR + document classification pipeline for automated digitisation and routing of incoming documents (invoices, delivery notes, customs forms)
– Custom text classification model (scikit-learn) trained on 1,200 labelled documents — achieved ~94% routing accuracy in production
– Integrated into the client's existing internal workflow; reduced manual document handling by ~70%
• ·
◦ ·