AI/ML Engineer focused on NLP, LLMs, RAG, Computer Vision, and applied AI products
I am an AI/ML engineer from Bishkek with experience in machine learning, NLP/LLM systems, computer vision, OCR, RAG, FastAPI, Docker, and data preparation. I recently graduated from Computer Engineering at Kyrgyz-Turkish Manas University and work on applied ML projects.
UlutSoft
Jan 2026 — PresentPart-time ML engineer working on applied ML and computer vision, including model training, evaluation, and iteration.
• Work on applied machine learning and computer vision tasks using Python, PyTorch, and image-processing tools. • Train, evaluate, and improve models; analyze outputs and iterate on model performance.
MBANK
A commercial bank in Kyrgyzstan.
Apr 2025 — Aug 2025Prepared ASR/TTS data and built a Kyrgyz text segmentation platform and document-based RAG prototype.
• Prepared dataset structure for ASR/TTS tasks, including audio segmentation logic and transcription format requirements. • Built a Kyrgyz text segmentation platform with FastAPI, TypeScript, Docker, and Apertium Kyrgyz. • Implemented a document-based RAG prototype using LangChain to answer questions from uploaded files. • Tested Docker-based deployment workflows for internal AI/NLP tools.
TÜBİTAK
Türkiye's government agency for supporting and coordinating scientific research and technological development.
Oct 2024 — Dec 2024Prepared Turkish and Kyrgyz datasets and preprocessing pipelines for multilingual NLP experiments.
• Cleaned and structured Turkish and Kyrgyz text datasets for NLP experiments. • Built preprocessing and tokenization pipelines for multilingual text data. • Tested multilingual text classification workflows and analyzed model outputs.
Kyrgyz-Turkish Manas University
2022 — 2026AI Academy KG
2024 — 2025Built a RAG system that answers user questions based on uploaded documents. Implemented document loading, chunking, embedding-based retrieval, and LLM response generation.
Generated 30,000+ synthetic OCR image-text pairs from Kyrgyz text using Genalog degradation profiles. Participated in VLM fine-tuning experiments and evaluated OCR quality using WER/CER. Contributed to an annotation workflow that processed 70,000+ images for dataset preparation.
Built an AI platform for helping applicants choose educational programs and prepare for exams. Worked on recommendation logic, AI assistant, tests, educational data processing, and backend APIs.
Built tabular ML models using data from mashina.kg and house.kg. Improved real estate price prediction to 9.9% MAPE through feature engineering and model optimization.
