Looking for a job

Aidar Amangeldi

@aidaramangeldi

AI Engineer at Medical Center of the President’s Affairs Administration of the Republic of Kazakhstan, building clinical LLM and RAG applications.

Kazakhstan

Message

At the Medical Center of the President’s Affairs Administration of the Republic of Kazakhstan, I develop LLM and RAG applications for physicians, pharmacists, and dietitians. I build Python services and clinical data pipelines that connect structured and unstructured sources for context-aware recommendations.

At Nazarbayev University’s ESAI Laboratory, I fine-tune and benchmark language and vision models for edge devices. I designed a GPU utilization-guided DVFS algorithm that reduced fine-tuning energy consumption by up to 37%, and I’ve co-authored two IEEE publications on energy-efficient deep learning and hardware-aware AI optimization.

Experience

Work history, roles, and key accomplishments

Current

AI Engineer

Current

Medical Center of the President’s Affairs Administration of

Apr 2025 - Present (1 year 7 months)

Developed enterprise LLM and RAG applications for clinical staff, building Python backend services, REST APIs, and vector-database-backed semantic retrieval. Built data pipelines integrating HL7 v2, FHIR R4, DICOM, and SQL sources, and containerized services with Docker and Kubernetes.

Python LLM RAG REST APIs Vector Databases Docker Kubernetes MLOps CI CD Weights Biases Tensorboard HL7 FHIR DICOM SQL

Current

Research Assistant – Embedded Systems & AI (ESAI) Laboratory

Current

Nazarbayev University

Aug 2022 - Present (4 years 3 months)

Fine-tuned and benchmarked transformer and CNN models for edge deployment, optimizing energy, latency, and accuracy. Designed a GPU DVFS algorithm reducing fine-tuning energy by up to 37% and co-authored two IEEE publications.

Python PyTorch Transformers BERT DeBERTa Pythia SmolLM CNN Vision Transformer NVIDIA Jetson DVFS Glue Pandas Matplotlib

Data Science Intern

Institute of Smart Systems and Artificial Intelligence (ISSA

May 2022 - Aug 2022 (3 months)

Contributed to the first Central Asian dishes dataset, web-scraping over 6,000 images and performing annotation. Trained YOLO and ResNet models achieving 80% classification accuracy and 0.83/0.80/0.87 precision/recall/F1.

Python Web Scraping YOLO ResNet Data Annotation Quality Control

Education

Degrees, certifications, and relevant coursework

Nazarbayev University

Master of Science, Data Science

2024 -

Pursuing a Master of Science in Data Science with a thesis on GPU utilization-guided dynamic frequency scaling for energy-efficient fine-tuning of large language models.