Skip to content

Данияр Есимов

@daniyar-esimov

ML / Computer Vision Engineer

ML / Computer Vision Engineer with production experience in real-ti me video and audio detection, document AI and data engineering.

Open to

Hiring & referralsProject collaborationResearch collaborationStartup collaborationJoining competitions & hackathons

About

Currently building an LLM-based document verification system at Freedom Finance I nsurance; previously delivered CCTV aggression detection, audio event detection and X-ray screening systems at Cerebro. Fourth-year Computer Science student at SDU University (B.Sc. 2027).

Career

Experience: 1 years

Experience

Data Scientist, AI Transformation Department

Freedom Finance Insurance JSC (Freedom Holding Corp.)

Kazakhstan insurer within Freedom Holding Corp., a financial services group with operations across Central Asia and the United States.

Jul 2026 — Present

Building an LLM-based document verification and archive system, with reported ~86% accuracy and ~0.89 macro-F1, plus RAG-based document recovery.

• Building an automated document verification and archive system on a locally hosted LLM (Qwen3.8-27B) with a hybrid VLM / PaddleOCR approach: classifies seven document types, extracts key fields and cross-checks them against 1C ERP records (~86% accuracy, ~0.89 macro-F1).

ML / Computer Vision Developer

Cerebro Innovation Technologies

Feb 2026 — Jul 2026

Built real-time CCTV and audio detection systems and an X-ray screening prototype, reporting event recall, false alarms, model accuracy and inference speed.

• Aggression detection: real-time CCTV fight detection on a fine-tuned X3D-S model; three-service Docker architecture (video capture, GPU inference, event dispatch) over Apache Kafka; live RTSP streams with per-camera zones and thresholds, webhook events.

Education

B.Sc., Computer Science

SDU University

2023 — 2027

Data Engineering

Freedom Holding Corp. & SDU University

2025 — 2026

Selected projects

Mastercard Text-to-SQL Chatbot

Natural-language analytics over transaction data, with a fine-tuned Gemma 4 4B model, a FastAPI and DuckDB backend, a React/TypeScript frontend in English, Russian and Kazakh, SQL-injection protection, and dynamic charts and tables.

NYC Taxi Fare Prediction Pipeline

End-to-end ML system built during the Freedom Data Engineering Program.

Achievements

1st place (team), Freedom Holding hackathon

Freedom Holding · 2026

Skills

PythonJavaSQL (PostgreSQL)LLM fine-tuning (LoRA / QLoRA / PEFTUnsloth)Prompt engineeringRAG (embeddingshybrid searchreranking)Vector DBs (QdrantChromaDB)LangChainLangGraphLlamaIndexMCPLocal LLMs / VLMsTransformersNERText classificationTokenization