Looking for a job

Arlan Toktarkhan

@arlantoktarkhan

Machine Learning Engineer at BTS Digital who raised OCR validation accuracy from 75% to 95% with a custom PyTorch architecture.

Kazakhstan

Message

At BTS Digital, I designed and trained a custom OCR architecture in PyTorch, raising validation accuracy from 75% to 95%. I also built a multi-GPU training pipeline and a synthetic-data generator that produces more than 500,000 labelled images.

I fine-tuned vision-language models for face anti-spoofing, reducing HTER from 5.0% to 1.3% on production validation data. I also exported models to ONNX Runtime and set up reproducible GitLab CI/CD promotion.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer

Current

BTS Digital

Mar 2026 - Present (8 months)

Designed and trained a custom OCR architecture in PyTorch, raising validation accuracy from 75% to 95%. Built multi-GPU training pipeline and synthetic data generator, and fine-tuned vision-language backbones for face anti-spoofing.

PyTorch Multi GPU Training Synthetic Data OCR CNN ViT RNN CTC SigLIP Perception Encoder Flip ONNX Runtime GitLab CI CD

Machine Learning Intern

BTS Digital

Sep 2025 - Feb 2026 (5 months)

Trained face detection and anti-spoofing models using RetinaFace and PyTorch, and ran transfer-learning experiments on deepfake video datasets to improve robustness.

PyTorch RetinaFace Transfer Learning Face Detection Antispoofing

Education

Degrees, certifications, and relevant coursework

University of London

Bachelor of Science, Data Science and Business Analytics

2023 - 2026

Grade: First Class Honours

BSc Data Science and Business Analytics with First Class Honours. Attended LSE Summer School, ME314 'Introduction to Data Science and Machine Learning', achieving grade A.

Kazakh-British Technical University

Bachelor of Science, Information Systems

2022 - 2026

Grade: GPA 3.79/4.0

Activities and societies: Teaching assistant for Calculus, Algebra and Statistics.

BSc Information Systems with GPA 3.79/4.0, highest in the International School of Economics. Served as a teaching assistant for Calculus, Algebra, and Statistics.