Abdulkhak Abdrakhman
I am a Machine Learning Engineer with over three years of experience delivering production-grade NLP and ML solutions.
About
I am a Machine Learning Engineer with over three years of experience delivering production-grade NLP and ML solutions. I have hands-on expertise in transformer fine-tuning using Hugging Face, vector search implementations with FAISS, and deploying FastAPI services. My passion lies in building scalable Retrieval-Augmented Generation systems and managing end-to-end MLOps workflows, especially in cloud environments.
Throughout my career, I have built LLM chatbots, sentiment classifiers, and optimization models, often utilizing Docker and MLflow for efficient deployment. I enjoy applying techniques like LoRA and distillation to optimize inference for limited compute scenarios and have a strong foundation in operations research to optimize business processes. I am open to relocation and eager to contribute to innovative AI-driven projects.
- 7 years' experience
- Joined September 2025
Skills· 5
Languages
- AfarAdvanced
Industry experience
- Software & Internet
- Manufacturing
- Telecommunications
- Education
More experts in Almaty, Kazakhstan
Work experience· 5 roles
Developed forecasting models that optimized inventory management, achieving an 80% reduction in material waste. Implemented operations research-based methods to streamline production scheduling and reduce overall cost structures. Seamlessly converted operations and processes into machine learning workflows, resulting in measurable financial benefits for the company.
Developed LangChain-based assistants integrating OpenAI and FAISS retrieval, deployed using FastAPI, Docker, and NGINX. Designed end-to-end Retrieval-Augmented Generation pipelines combining structured SQL data with unstructured PDFs to create efficient enterprise search systems. Fine-tuned transformer-based sentiment classifiers on custom domain datasets using Hugging Face and PyTorch frameworks. Applied Low-Rank Adaptation (LoRA) and knowledge distillation methods to optimize transformer inference for environments with limited computational resources.
Created a DistilBERT-based emoji-sentiment analysis pipeline for real-time emotion tracking in chat applications. Designed fuzzy-logic multi-criteria decision models for robust scoring systems. Conducted fine-tuning of transformer models on low-resource datasets, ensuring reproducibility of experiments using sklearn and Hugging Face transformer libraries.
Education
Hire Abdulkhak Abdrakhman today
Post a job and invite Abdulkhak, or contact them directly. No hidden commission.