Abylai Kairkhanov

AI Engineer

DAN and PARTNERS

Architecture and design, construction and real estate.

Built and deployed AWS based RAG and GraphRAG systems for construction code audits, optimizing local model inference, routing, and production monitoring.

• Designed and deployed a production-grade, multi-stage RAG pipeline on AWS to automate compliance checks against Kazakhstan construction codes. Integrated an automatic ContextQualityGate filter with a minimum threshold of 0.15 and custom citation repair mechanisms. • Optimized a local inference stack of about 28.1 billion parameters on an AWS EC2 g6e.4xlarge instance with an NVIDIA L40S GPU and 128 GB RAM. Used vLLM with FP8 quantization for Qwen3.6-27B-FP8 and HuggingFace TEI for BGE-M3 and BGE-Reranker-v2-m3, increasing token throughput and reducing hosting costs.

AI Engineer

FTT AI

Fine-tuned and served LLMs, and built RAG text analytics for over 50,000 documents, with reported gains in accuracy, latency, and retrieval.

• Fine-tuned Qwen 3.5 9B using 4-bit QLoRA on a specialized corporate dataset, reducing training VRAM requirements by 60 percent so training could run on one GPU, and improving target domain answer accuracy by 15 percent. • Designed a text analytics platform using a multi-stage RAG pipeline, with multilingual-e5-large 1024-dimensional embeddings, a Qdrant HNSW graph, and a bge-reranker-large cross-encoder. Enabled automated semantic analysis of more than 50,000 unstructured documents.

Software Engineer

First Credit Bureau

Financial sector, including credit and financial services.

Developed a Go tool for XML comparison with reporting and integrations, and made large file checks up to five times faster.

• Developed a Go application to compare XML files automatically and record structural differences. • Implemented change notifications and PDF and Excel report generation with email delivery. • Added a REST API and CLI for integration and ease of use, and documented the API contract in OpenAPI. • Set up CI/CD with GitHub Actions, Docker containerization, and tests covering about 80 percent of critical logic. • Optimized the comparison algorithm, making checks of large XML files up to five times faster than the original solution.

Software Engineer

QUSH

Startup developing an online service for booking sports courts and fields.

Helped build a sports venue booking MVP with Python, React, PostgreSQL, and Docker. The startup reached the final of a student research competition.

• Worked with a team to develop an MVP for QUSH, an online sports court and field booking service, integrating business logic and analytics. • Built the backend with Python FastAPI and PostgreSQL. • Developed the React frontend for finding venues, selecting times, paying, and managing bookings. • Set up deployment and Docker containerization. • Integrated a business plan and analytics modules to track venue occupancy and service performance. • The startup reached the final of a student research competition.

Bachelor's degree, Information Technology, Software Engineering

Astana IT University

Higher education, Reserve officer, geographic information systems

Military Department