Yerassyl Pirzhanov

@x0r19

Agentic AI Engineer / AI Lead

Agentic AI Engineer / AI Lead with production experience building LangGraph-based LLM agents, multi-step agentic workflows, RAG pipelines, semantic search, graph analytics, anomaly detection, OCR, APIs, and backend infrastructure for regulated financial and public-sector environments.

About

Strong Python engineer with hands-on experience designing reliable AI systems over complex internal data: tool calling, ReAct-style execution, retrieval evaluation, prompt/output validation, observability, latency analysis, AI safety controls, and human-in-the-loop review. Strong product sense with experience translating ambiguous stakeholder needs into production systems through rapid prototyping, measurable iteration, deployment, and close collaboration with technical, domain, and client-facing teams.

Career

Experience

Head Expert / AI Engineer, Project Lead

Financial Monitoring Agency of the Republic of Kazakhstan

Jan 2025 — Present

Led end-to-end delivery of production AI systems, from architecture and backend implementation to deployment and observability. Built LLM/RAG and agentic solutions using PostgreSQL, Elasticsearch, Qdrant, Neo4j, OCR, and secure tool integrations. Improved retrieval, graph analytics, anomaly detection, reliability, and human-in-the-loop workflows, while leading key AI initiatives from prototype to production.

Python Developer

Tech Tailors

Aug 2024 — Jan 2025

Built backend services and REST APIs for data-driven products using Python, Django/FastAPI-style service patterns, relational databases, and clean API boundaries. Improved service performance and maintainability through profiling, refactoring, query optimization, and clearer API contracts.

Data Scientist / ML Automation Developer

Research Center “TALAP”

Jan 2023 — Nov 2024

Analyzed multi-terabyte tax datasets for public-sector research using Python, pandas, NumPy, and Dask for cleaning, transformation, and exploratory analysis. Built automation tools for news aggregation, tagging, summarization, and search; expanded the MVP after a successful pilot. Designed a multi-level classifier for macro/micro topic tagging and integrated LLM APIs and embedding-based semantic search for prompt-driven retrieval.

Education

Big Data Analysis

AITU

2021 — 2024

B.Sc. in Informatics, Big Data Analysis. Curse work on building GAN, for anomaly detection on video images.

Selected projects

AQIQAT

Led development of AQIQAT LLM workflows using RAG over construction estimates and reference-price sources, HyDE-assisted retrieval, tool-based calculation flows, grounded explanations, and iterative validation with domain experts. The agency publicly reported that AQIQAT helped prevent 242.4B KZT in unnecessary spend in 2025, including 30B KZT on the Comfort School program.

Open project