Hello friend!

✨ About Me

I’m a 31-year-old data scientist from Kazakhstan, currently based in Wrocław, Poland. I hold a Bachelor’s degree in Mathematics with honors from the Kazakhstan Department of Moscow State University (named after M.V. Lomonosov). Toward the end of my undergraduate studies, I discovered my passion for Data Science after completing my first course in the field. That interest evolved into a career when I won an Olympiad that secured my admission to the Master’s program in Computer Science (Data Science track) at HSE University in Moscow.

Throughout my academic journey, I gravitated toward the theoretical aspects of data science, particularly those intersecting with complex data structures such as graphs. This aligned with my research work at the IITP RAS lab “Data Analysis in Neurosciences,” where I explored graph-based generative models and brain MRI analysis. My academic efforts led to presentations at top-tier conferences like MICCAI.

Eager to translate my research skills into real-world applications and because of burnout from the science work, I transitioned into the commercial sector. At Constanta (was OSAI AI, now part of SportRadar), I engineered computer vision systems for real-time video segmentation, object tracking, and player recognition in both esports and basketball contexts. I then joined the startup Halbestunde, where I not only tackled audio and visual recognition tasks but also led the implementation of development workflows and infrastructure — a unique opportunity to blend engineering, planning, and product thinking.

Later, at VK’s voice assistant team Marusya, I worked on audio classification and segmentation, optimizing internal tools and improving evaluation metrics. Lately, I worked as a Senior Data Scientist at Akvelon, focusing on large language model (LLM) applications, MLOps, and intelligent assistant development using RAG, Langchain, and ReAct frameworks. I’ve also introduced performance monitoring systems (Grafana, Prometheus) and developed prompt optimization pipelines.

Latest Position (2025)

Principal Machine Learning Engineer: Anecdote AI — Remote

At Anecdote AI, I lead the development of core features that power agentic, LLM-based research workflows:

• 🚀 Deep Search: Designed, implemented, and shipped a state-of-the-art agentic search system delivering 40% more accurate summaries compared to the baseline. I managed the full lifecycle from architecture to deployment.
• 📊 Quantify: Built the Quantify feature — an automated system for extracting and quantifying structured insights from natural language data. Developed and maintained both integration and end-to-end test coverage to ensure robustness.
• 🔍 RAG System Optimization: Managed the Retrieval-Augmented Generation (RAG) pipeline to enhance context retrieval accuracy — improving performance on 80% of evaluation cases.
• 🧪 Testing Infrastructure: Set up and maintained comprehensive test suites (integration + E2E) for key components, ensuring high reliability in production.

This role has combined technical leadership, LLM engineering, and deep product understanding — all in a fast-paced, high-impact startup environment.

Across both academia and industry, I enjoy bridging theory with application — whether in vision, audio, or language — and am passionate about building intelligent systems that solve real-world problems. 💡🤝

I want to try to make an honest blog to share the path in the industry. I’m not ultra smart or a hard worker. I’m smarter than average, as shown by my achievements like both Diplomas with Honors. Also, I found out I belong to the Autistic Spectrum Disorder which helps and overwhelms me at the same time in my work. There’s so much more to discover about myself. It is frustrating and exciting at the same time.

About my projects

• Commercial projects
  • Automatic systems for cybersports: Segmentation and OCR of information in-game
  • Automatic tracking, detection and recognition of players in physical basketball
  • Audio classification of noisy and denoising (Neurolab test task)
  • Notes recognition: from photos, pdfs, etc
  • Classification of ID cards (test task for Revolut, got an offer)
  • Detection of mistakes in audio with reference
  • Endpoint detector in phrases to the fast replay of voice assistant
  • Chatbot for Lawyers
  • Anecdote AI: Created Deep search experience from scratch
  • Implemented Quantify tool from scratch
• Academic
  • Optimization of parameters for LDDMM registration model task base:
    • Straightforward approach
    • Template approach
  • Diffeomorphical Transformer (like Spatial Transformer Network)

Skills

Python, PyTorch, PyCharm, Docker, Make, Bash, DVC, Hydra, HDFS, LLM, Langchain, Elastic Search, Hugging Face, Transformers, etc. see my CV