Azamat Kozhakov

AI engineer

Qalan.kz

AI engineer at Qalan.kz • Qalan.kz, Almaty, Kazakhstan • Redesigned a legacy rule-based chatbot into a scalable Multi-Agent orchestration system using LangGraph, serving 70,000 daily users across 3 regions (KZ, RU, UZ) • Automated 79% of total customer support chats by implementing LLM-driven agents, intelligently routing the remaining complex queries (complaints, bug reports) to human operators. • Reduced AI inference spend by 86% by identifying and eliminating a legacy bug generating phantom backend requests, introducing prompt caching across eligible agents, and routing suitab

Lead AI Architect

Agartu Academy Project(Qalan.kz)

Agartu Academy (by Qalan.kz), Almaty, Kazakhstan • Architected a sophisticated multi-role RAG system using the native Vertex RAG engine, agent platforms, and Vector Search. Implemented dynamic routing between Gemini and ChatGPT to optimize responses based on specific domain roles and distinct knowledge bases. • Integrated Document AI for complex OCR tasks to feed high-quality contextual data into the vector storage. • Developed an innovative multimodal pronunciation assessment service (Tajweed) utilizing Gemini’s Native Audio Processing (NAP), bypassing traditional STT latency f

BSc Computer Science

SDU University