Looking for a job

Erdauit Torekhan

@erdauittorekhan

Message

Computer vision researcher at Qoima Automation who built real-time analytics for 15 camera streams on one machine.

Kazakhstan

Message

At Qoima Automation, I designed and implemented a computer vision system for multi-camera analytics, processing 15 camera streams on a single machine. I optimized inference with TensorRT, quantization, and NVIDIA Triton, reducing latency roughly 5–6x.

At OZ Forensics, I designed and trained a lightweight LSTM model for face presentation-attack detection. My work also includes building synthetic training-data and RAG pipelines, and researching visual features for a first-author IEEE SIST 2025 study.

Experience

Work history, roles, and key accomplishments

Current

Lead Computer Vision Researcher

Current

Qoima Automation

Apr 2026 - Present (7 months)

Designed and implemented the computer vision architecture from scratch, including detection, multi-object tracking, and vision-language models. Optimized inference with TensorRT and NVIDIA Triton, cutting latency 5-6x, and built the full training pipeline with CI/CD.

Computer Vision Object Detection Multi Object Tracking Vision Language Models TensorRT NVIDIA Triton Inference Server FFMPEG YOLOv8 SAM CLIP Dino Weights Biases CI CD CVAT Docker Kubernetes

Current

Research Engineer

Current

FLIS Lab, Kazakh-British Technical University

Feb 2024 - Present (2 years 9 months)

Led visual-feature extraction and evaluation for a first-author study on painting aesthetics, published at IEEE SIST 2025. Contributed to research on multimodal emotion recognition and fuzzy color representations, and mentored undergraduate researchers.

Feature Extraction Emotion Recognition Fuzzy Color Representations Research Mentorship

Middle Computer Vision Researcher

OZ Forensics

Aug 2025 - Apr 2026 (8 months)

Worked on face presentation-attack detection for an identity-verification platform handling 100M+ verifications per month. Designed and trained an original lightweight LSTM-based model, benchmarked with ONNX Runtime and TensorRT, and built threshold-aware evaluation tooling.

Antispoofing LSTM ONNX Runtime TensorRT FP16 INT8 Quantization GCP Weights Biases FFMPEG FAR Operating Point Analysis

Junior Computer Vision Engineer

BTS Digital

Sep 2024 - Aug 2025 (11 months)

Built the team's first synthetic training-data generation pipeline with ComfyUI and developed face presentation-attack detection models, tracking metrics under ISO/IEC 30107-3. Also built a domain-specific RAG system using LangChain and vector search.

Synthetic Data Generation ComfyUI Antispoofing APCER BPCER Acer RAG Lang Chain Vector Search Deep Seek

Data Engineer

Adata

Apr 2024 - Sep 2024 (5 months)

Developed ETL pipelines with Talend and Python, refactored Java/Python processing bottlenecks, and implemented pandas-based data cleansing and validation.

ETL Talend Python Java Pandas Data Cleaning Data Validation