ML engineer
ML engineer that want to solve high impact real world problems. Hate doing useless work
National Analytical Center
Jul 2025 — PresentOwns the AI component of a content analysis project for government entities and increased offline LLM generation throughput 2.5x.
• Implemented offline LLM inferencing, achieving 2.5x LLM generation throughput compared to the existing implementation using KV caching and concurrent batch processing.
Kazakhtelecom
Sep 2025 — Mar 2026Worked on an AI procurement price-analysis platform that secured pre-rollout contracts and deployed a 24/7 Q&A assistant for the Samruk-Kazyna Trust website.
• Responsible for the technical implementation of an AI-driven marketing price analysis platform intended to prevent price manipulation in procurement and support objective, bias-free market pricing. The system translates regulatory requirements into solutions and removes human intervention to identify the market value of goods. Secured pre-rollout contracts with multiple major national enterprises. • Developed and deployed a 24/7 Q&A AI assistant for the Samruk-Kazyna Trust fund website (sk-trust.kz) to answer potential applicant and general questions about the fund.
NITEC
Mar 2025 — Apr 2025Evaluated embedding models for Egov.AI, improving Russian and Kazakh top-3 retrieval accuracy by 46%, and built on-premise AI prototypes for presentations and meeting reports.
• Evaluated open-source embedding models for Egov.AI on production benchmarks and selected one that improved Russian and Kazakh top-3 retrieval accuracy by 46%. • Built an MVP of an on-premise AI slide generator using LLMs and text-to-image models. • Developed an audio-to-report pipeline for government use, using on-premise transcription models and LLMs to summarize meetings and reduce manual documentation effort. • Contributed to the KazLaw project by segmenting legislative texts into semantic blocks to improve downstream information-retrieval accuracy.
ReLive Intelligence
Apr 2024 — Jun 2024Built computer-vision models and FastAPI endpoints for facility safety, fall detection, and protective-equipment monitoring.
• Developed deep-learning models for object detection and image classification, managing development from data collection through model training and performance analysis. • Built systems for water-spill detection, human-fall detection, and personal protective equipment violation monitoring. • Implemented APIs using FastAPI.
Astana IT University
2022 — 2025Middle East Techincal University
2024 — 2025A retrieval-augmented generation chatbot for Middle East Technical University, using a knowledge base extracted from metu.edu.tr subdomains to answer questions about the university.
Open projectA computer-vision system to detect people and assess compliance with safety helmet and jacket requirements, implemented as a FastAPI API.
Open projectCompared dense retrieval models with BM25 and experimented with fine-tuning embedding models as part of a Web Search Engine Design course.
Open project2024
