Anatolii Chetrafil, Computer Vision Lead. AI Model Audit

Languages

English
Native or bilingual
Norwegian
Conversational
Swedish
Basic
Russian
Fluent
Ukrainian
Fluent
Bulgarian
Basic

Workplace preferences

Remote only

Primarily works remotely

TaskFlock
AI Model Audit & Computer Vision Lead (Founder)
E-COMMERCE
January 2024 - Today (2 years and 5 months)
Oslo, Norway
TaskFlock — a CV/data annotation team. I lead delivery from data → model → edge deployment.

What I do:
• 72-hour AI Model Audit: FP/FN error map, one acceptance rule (metric+threshold), 3–5 quick wins, mini demo.
• Monthly Model Care: hard-case mining, re-training and releases (1–2/mo), “errors-to-watch” report.
• SKU Onboarding ($99): photo/video pack, annotation (bbox/track/events), micro-calibration, test clip (24–48h).

Stack & methods:
• Labeling: CVAT, Label Studio, Roboflow (bbox, polygon, masks, keypoints, tracking). QC: double pass, IAA ≥95%.
• Data formats: COCO, YOLO, Pascal VOC; video chunks, event timelines.
• Training: PyTorch, YOLOv5/v8, Detectron2; metrics mAP/F1; augmentations.
• Edge/opt.: ONNX, TensorRT, quantization (FP16/INT8), Jetson/CPU, RTSP pipelines.
• Security: NDA/DPA, work in client VPC/on-prem. White-label friendly.

Results:
• Retail/CCTV pilots: −30–40% false alarms, +10–25% accuracy lift; weekly releases.
• Team capacity: 6 annotators + QA lead (100k+ labels/day peak).

Available for: 72-hour Audit • Monthly Model Care • SKU Onboarding • Starter Packs (retail/CCTV/smart devices).

Be the first to recommend Anatolii

Help this freelancer shine by sharing your experience working together.

A

Agatha Frydrych

Backend Java Software Engineer

4.7

(3)

2

B

Baptiste Duhen

Fullstack developer

4.6

(4)

5

A

Amed Hamou

Senior Lead Developer

4

(2)

7

A

Audrey Champion

Web developer

4.3

(3)

4

Signup to reveal