Ilnar Shafigullin
Staff Data Scientist @Sberbank
Signup · Get unlimited contacts
WORK HISTORY
Staff Data Scientist @Sberbank
Built and productionized a Document AI pipeline for credit/insurance - OCR (Surya, Tesseract)-> document classification & clustering (DiT/BERT + TF‑IDF/embeddings)-> vector search for retrieval/matching; at target precision > 0.95 achieved recall ≈ 0.5; automation reduced case handling from weeks to minutes- Developed long‑document NLP and topic clustering: hierarchical chunking -> chunk embeddings -> attention/aggregation; fine‑tuned BERT‑like models via LoRA- Shipped models to production and integrated with existing systems: Django APIs, Docker containers, Kubernetes orchestration; implemented monitoring/logging and scheduled re‑training pipelines- Designed workflow‑automation agents using LangChain and Model Context Protocol (MCP) to orchestrate tools/APIs and RAG pipelines based on embedding‑driven vector search; implemented monitoring/logging and human‑in‑the‑loop fallbacksTech Stack- Languages & Libraries: Python (pandas, numpy, scikit‑learn), PyTorch- Models: BERT‑family (incl. LoRA), DiT/ViT; TF‑IDF, embeddings- OCR: Surya, Tesseract- Agents & Orchestration: LangChain, MCP (Model Context Protocol)- Clustering: HDBSCAN (silhouette evaluation)- Retrieval: Embedding‑based vector search- Deployment & Integration: Django, Docker, Kubernetes- Validation Schemes: Train/Validation/Test; time‑based cross‑validation where applicable
EDUCATION
Kazan State University
Specialist, Mathematics
Innopolis University
Specialist, Data science
Kazan State University
Candidate of Physical and Mathematical Sciences, Mathematics
Kazan State University
Specialist, Business Administration and Management, General
ABOUT ILNAR SHAFIGULLIN
Staff Data Scientist / ML Engineer building end-to-end ML systems: problem framing…
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.