Looking for a job

dmytro stepanchuk

@dmytrostepanchuk

ML Engineer at Kyivstar.Tech building post-training methods for a national LLM and Ukrainian-language data quality models.

Ukraine

Message

At Kyivstar.Tech, I designed methodologies for high-throughput synthetic data generation and natural-data translation for national LLM training. I also built a model to assess pre-training data quality and detect pro-Russian narratives in Ukrainian-language data.

At PrivatBank, I rebuilt and optimized a customer propensity model, developed customer segmentations and contributed to a multi-year LTV forecasting framework. I also built a multi-agent system that reduced a workflow from 2–3 days to under 20 minutes.

In the Kaggle competition “Handwritten to Data,” I built a multi-stage Ukrainian handwriting recognition pipeline and fine-tuned vision-language models; the work took 1st place on the public leaderboard. I also teach AI topics, including computer vision and LLM benchmarking.

Experience

Work history, roles, and key accomplishments

Current

Lecturer

Current

Kyiv School of Economics

Oct 2026 - Present (1 month)

Architecting the university-wide AI adoption strategy and teaching the Advanced Computer Vision course. Faculty lecturer at the Agentic AI School, covering advanced topics.

AI Strategy Computer Vision Teaching Agentic AI

Current

ML Engineer (NLP)

Current

Kyivstar.Tech

Jan 2026 - Present (10 months)

Designed methodologies for high-throughput synthetic data generation and translation pipelines for national LLM training. Built a SOTA model for pre-training data quality assessment and detection of pro-Russian narratives, and developed alignment methodology for post-training.

LLM NLP Synthetic Data Generation VLLM Model Serving Inference Optimization Data Quality Assessment

Data Scientist

PrivatBank

Aug 2024 - Jan 2026 (1 year 5 months)

Rebuilt and optimized a legacy customer propensity model, built customer segmentations, and developed a multi-year LTV forecasting framework. Delivered generative AI solutions and a multi-agent system that reduced workflow time from 2-3 days to under 20 minutes.

Data Science Customer Analytics Propensity Modeling Segmentation LTV Forecasting Generative AI Multi Agent Systems Python SQL

CV Engineer

NDA

Mar 2024 - Aug 2024 (5 months)

Built end-to-end computer vision solutions for classification and object detection, including synthetic dataset generation and low-latency API. Developed self-hosted generative AI for payment card designs and OCR+LLM document processing with graph-augmented RAG.

Computer Vision Object Detection Synthetic Data Generative AI OCR LLM RAG fastAPI

Machine Learning Engineer

Postdata AI

Mar 2022 - Aug 2024 (2 years 5 months)

Worked on client projects involving Python codebase optimization, SQL performance, database design, and applied ML solutions. Built dashboards and forecasting tools, and automated workflows with GCP and FastAPI.

Python SQL Database Design Machine Learning Computer Vision regression GCP fastAPI Dashboarding Forecasting

Software Engineer

Self-employed

Aug 2021 - May 2022 (9 months)

Developed an automatic scheduling utility using constraint-satisfaction and optimization, a semi-automatic file sorting application, and a CMS website for a local media team.

Constraint Satisfaction Optimization GUI Development Document Processing Image Processing Fullstack Development CMS