Aspandiyar Nurimanov

ML Engineer

Skyro

Built and productionized customer-conversation ML pipelines, including intent classification for 940K+ threads at 87% F1 and a local model that avoided about $10K/year in API costs.

• Built an end-to-end customer-contact intent classification pipeline processing 940K+ support threads YTD across a 12 top-level / 85 canonical intent taxonomy, designing the taxonomy and preparing all training data in-house and reaching 87% F1 accuracy; powers company-wide visibility into contact drivers.

Research Data Scientist

Institute of Smart Systems and Artificial Intelligence (ISSAI)

Built LLM data and inference pipelines and deployed a RAG system for low-resource-language retrieval; co-authored two listed publications.

• Engineered a scalable data pipeline to collect, preprocess, generate, and translate datasets for SFT and large-scale language model training; optimized multi-GPU LLM inference on A100/H100 clusters for high throughput and low latency. • Deployed a production-grade Retrieval-Augmented Generation (RAG) system leveraging Agentic AI for low-resource-language knowledge retrieval.

AI/ML Engineer

BennuData

Applied embeddings and clustering to AWS cloud discovery and built BERT/GPT-3 models to automate Disaster Recovery Plan generation.

• Conducted AWS cloud discovery using AI-powered BERT embeddings, then clustered and classified resources into logical applications (K-Means, Self-Organizing Maps) to streamline architecture analysis for the company's SkyDeck-backed product. • Developed and fine-tuned BERT and GPT-3 models to automatically generate Disaster Recovery Plans, reducing manual planning time.

Senior Data Scientist

Kumtor Gold Company

Gold mining company in Kyrgyzstan.

Built and deployed an incident-ticket classifier reported at 98% accuracy and designed digital twins of production data flows.

• Developed an NLP model to classify incident tickets with 98% accuracy; deployed as a Flask API in Docker on Kubernetes for scalable, production-grade inference. • Designed digital twins of company production-cycle data flows, enhancing process monitoring.

Data Scientist

KPMG

Professional services firm.

Automated tender analysis, saving 5.5 staff-hours per day, and delivered AI solutions and training for client organizations.

• Automated tender parsing and built an NLP-based tender classification and sentiment analysis system, saving 5.5 staff-hours/day; deployed AI solutions for geotechnologists at a nuclear fuel cycle client. • Delivered “Data Science for Managers,” “Power BI,” and “Data Management” webinars for Kazakhstan ministries and corporations.

MSc, Applied Data Science

Frankfurt School of Finance & Management

BSc, Economics and Finance

University of London (LSE) / KBTU