M.Sc. AI & Data Science — Paris Dauphine University – PSL
I build AI systems that work in production — multi-agent pipelines, RAG architectures, and LLM-powered workflows. My background spans both applied engineering (shipping agents to production) and research (adversarial training, generative models, NLP for low-resource languages, computer vision).
Graduated from Paris Dauphine University – PSL Research University (top 20 globally), where I studied under researchers pushing the boundaries of ML theory. Before that, CS at University of Strasbourg. Five years living and working in France, now based in Almaty.
AI / ML
LangChain RAG LLM APIs Prompt Engineering LLM Evaluation PyTorch TensorFlow Hugging Face Adversarial Training GANs NLP ASR
Languages & Frameworks
Python FastAPI SQL JavaScript C/C++
Infrastructure
Docker Linux Git ChromaDB REST APIs
Research Methods
Model Benchmarking LLM-as-a-Judge Time-Series Analysis
AI Engineering Intern — MYTEAM.ai (Paris, Apr–Oct 2025) Built multi-agent LLM systems (LangChain + custom orchestration) deployed to production via FastAPI. Designed a RAG pipeline with Mistral OCR + ChromaDB for document intelligence. Developed a Gemini-powered prompting pipeline that reduced report generation time by ~60%. Created an evaluation framework combining LLM-as-a-Judge with Ridge regression, benchmarked via MAUVE scores. This was also my Master's thesis.
ASR Research Collaboration — Verint Systems (Remote, Dec 2025–Jan 2026) Data preparation and quality analysis for an Automatic Speech Recognition pipeline. Led technology selection and literature review.
Research Assistant — INRIA (Strasbourg, Jan–May 2024) Analysed high-dimensional electrophysiological time-series from neuroscience experiments using PCA, ICA, and spectral analysis to study cortical learning.
| Project | What it does | Key result |
|---|---|---|
| Robust Neural Networks | DeepFool as both attack & defense mechanism; Boosted Adversarial Training | 50% robustness vs PGD attacks (baseline: 16%) |
| Hybrid Recommendation System | BERT embeddings + Matrix Factorisation | RMSE −9.7%, training 52% faster |
| Legal Case Search — Hackathon - French Team 🥉 | LegalBERT + Llama for cross-border EU case retrieval | 3rd place, 6 EU countries, partnership talks with European Council |
| Kazakh QA Dataset | 15K row open-source QA dataset for Kazakh NLP | Public dataset for low-resource language research |
| GAN Image Generation | Custom GAN architecture | FID score 54, 15 min training on RTX 3060 |
- Agentic AI — multi-agent coordination, tool use, long-horizon reasoning
- LLM Evaluation — reliable metrics, LLM-as-a-Judge, calibration
- Low-resource NLP — Kazakh language, multilingual models
- Generative Models — GANs, diffusion, creative applications
M.Sc. AI & Data Science — Paris Dauphine University – PSL (#19 globally) · Graduated Dec 2025
B.Sc. Computer Science — University of Strasbourg · 2020–2023 (Top 10%)
🇬🇧 English — Proficient · 🇫🇷 French — Proficient · 🇷🇺 Russian — Native · 🇰🇿 Kazakh — Native/Conversational · 🇰🇷 Korean - Conversational
Open to AI/ML Engineer and Research Engineer roles — remote or relocation