Almat Kairbek

Middle Data Scientist

Kazakhtelecom JSC

Kazakhstan's national telecommunications operator.

Built NLP classification and ranking pipelines, monitored LLM outputs, engineered data marts, and developed an OFD analytics platform.

• Built an NLP text classification model (500-class BNS receipt categorizer) for a banking client, reaching Precision/Recall of 0.70–0.75; owned the lifecycle of feature/text preprocessing • Monitored and corrected outputs of an LLM-based agent generating regulatory legal documents, building a systematic error-review process to reduce hallucinated/incorrect content

Graduate Researcher – Semantic Search / Information Retrieval

Kazakh-British Technical University

A technical university in Kazakhstan.

Compared sparse, dense, and hybrid retrieval methods and built an unsupervised NLP pipeline for identifying IT trends; the retrieval study was published.

• Comparative study of sparse vs. dense retrieval (BM25 vs. Dense) and hybrid RRF fusion on a Wikipedia-derived NLP corpus, evaluated on nDCG@10 – core semantic-similarity/ranking methodology; published in Computación y Sistemas (Q4, Vol. 30 No. 1, 2026) • Master’s thesis: IT-trend identification via semantic analysis of publication databases using BERTopic, SPECTER v2 embeddings, UMAP, HDBSCAN, and custom topic-coherence metrics – end-to-end unsupervised NLP pipeline from embedding to clustering to interpretation;

Master of Science, Data Science

Kazakh-British Technical University

Bachelor of Science, ICT (Big Data Analysis)

Astana IT University