Skip to content

Alima Bazenova

@bazenovaaConnected with DSML. This person's Telegram account has been confirmed through DSML Kazakhstan.

Machine Learning Engineer

AI/ML engineer with hands-on experience in computer vision, data preparation, model training, evaluation, and end-to-end pipeline development. Enjoy learning new things!

Open to

Finding a mentorJoining pet projectsPaper discussionsHelping organize meetupsResearch collaboration

Career

Experience: 1 years

Experience

Research Assistant – AI for Clinical CT Imaging

Nazarbayev University

Research university in Kazakhstan.

Aug 2026 — Present

Built a CT-based pipeline for osteoporosis and sarcopenia screening in 280 patients; automated vertebra and psoas segmentation with mean Dice 0.89 against expert-corrected masks.

• Developed a deep-learning pipeline for opportunistic osteoporosis and sarcopenia screening from routine lumbar-spine CT, linking L3 psoas muscle biomarkers to DXA bone mineral density in 280 patients. • Automated vertebra and psoas segmentation on a multi-vendor cohort (5 manufacturers): baseline mean Dice 0.89 vs expert-corrected masks; the validated annotation set is used to fine-tune the model and train CNN classifiers for osteoporosis prediction.

Research Assistant – Automated Microscopy & Image Analysis

Eötvös Loránd University

Research university in Hungary, also known as ELTE.

Sep 2025 — Jul 2026

Automated microscopy and image-analysis workflows, released a pollen detection dataset, and benchmarked detectors with leakage-free validation; also trained a chironomid classifier on 27,774 images.

• Engineered an automated microscope-scanning and live-detection workflow replacing manual slide counting (75% faster); relative abundances matched an expert palynologist on 3 fully independent slides (Pearson r = 0.86, Spearman ρ = 0.80). • Built and released a 22-class detection dataset (3,001 images, 8,460 boxes) and benchmarked 7 CNN/transformer detectors (YOLOv8–v12, YOLO26, RT-DETR; 21 runs) under a leakage-free depth-based split: best YOLOv12l reached 0.496 pollen mAP@0.5, while a random split inflated it to 0.749 (+51%).

Junior ML Engineer – Video Behaviour Recognition

Green Eye Ltd

May 2025 — Aug 2025

Built a roughly 20,000-frame/clip behaviour dataset and benchmarked computer-vision approaches and transfer-learning baselines across 15+ configurations.

• Built a 6-class behaviour dataset (∼20K labelled frames/clips); benchmarked 10+ CV approaches and pretrained-CNN transfer-learning baselines (15+ configurations) by precision, recall, macro F1 and confusion matrices.

Education

MSc, Environmental Sciences

Eötvös Loránd University

2024 — 2026

BSc, Biological Sciences

Nazarbayev University

2020 — 2024

Selected projects

Safe Camera – Multi-Stream Video Behaviour Recognition

Fine-tuned a Kinetics-400 pretrained R3D-18 on RWF-2000, achieving 74.5% accuracy, 80.5% recall, and 75.9% F1. K-of-N temporal smoothing with a calibrated threshold reduced false detections fourfold at fixed recall. Processed 8 synchronized streams at 25 FPS with 0.2% frame loss and 176 ms p95 latency; reduced inference cost 3.8× with accuracy-neutral results by McNemar test. ONNX Runtime and i...

Health Passport AI

Built a Gemini-API pipeline for parsing PDF, DOCX, and image medical records, translating them, and flagging abnormal lab values at the Big Berlin Hackathon.

Skills

Deep learning for image/video (PyTorch)Object detectionCT segmentationPose estimationAction recognitionDataset construction & annotation QCReal-time multi-stream videoStatistical testing (ANOVA, non-parametric)Inter-observer agreementLeakage-free validationPrecision/recall/F1/mAP/DicePythonNumPypandasOpenCVSimpleITKnibabelGitFastAPI