Machine Learning
- — PyTorch
- — HuggingFace Transformers
- — scikit-learn
- — NumPy
- — pandas
Dmitry Dydin — Machine Learning · NLP
Fitpex.
Machine Learning & NLP Engineer
I build ML models and NLP systems — fine-tuning transformers (BERT/RuBERT), RAG pipelines and end-to-end ML workflows, with a background in information security.
Capabilities
Skills & tools
Machine learning and NLP engineering, with a security mindset.
NLP / LLM
- — spaCy
- — NLTK
- — LangChain
- — RAG pipelines
- — fine-tuning (BERT/RuBERT)
Tooling
- — Python
- — C++
- — Git
- — Docker
- — FastAPI
- — Linux
Computer Vision
- — OpenCV
- — YOLOv8
- — torchvision
- — ONNX
Security
A background in information security — analysing binaries and web pages for threats.
Education
Stankin MSTU — Applied Informatics · Rostelecom: Data Analytics & AI
Russian — native · English — B2
Recognition
Awards & competitions
Selected results from hackathons and research contests.
1st place
HackInHome 2025
with VK · ML + Frontend
1st place
Eureka SPC
Programming & Computer Engineering
2nd-degree laureate
All-Russian NTT contest
PoW blockchain implementation
Finalist PurpleHack 2026DataFusion 2026DS ExpeditionLCT 2025Alfa-Future 2025
Projects
Selected projects
ML and NLP projects from hackathons and research.
01 02 03 04
Bank review sentiment analyzer
Multi-class sentiment classification of bank reviews by subcategory. Fine-tuned RuBERT to 88% accuracy on 15,000 test reviews.
PyTorch · RuBERT · TransformersVoice emotion recognition
Real-time classification of 7 emotional states from audio (81% accuracy), integrated into a speech-to-text pipeline to enrich transcripts.
PyTorch · librosa · torchaudioML vulnerability scanner
A tool that analyses binaries and web pages for security threats using graph ML. Work in progress.
PyTorch Geometric · scikit-learnReceipt reconciliation system
Automated extraction of structured data from PDF receipts (parsing + OCR), receipt-type classification and reconciliation against the POS system to flag discrepancies.
PyMuPDF · scikit-learn · PyTorch Contact
Let's Connect.
Feel free to reach out through any of these platforms. I'll get back to you as soon as possible.
fitpex.m@gmail.com