Ian Shmidt

ML Engineer

REDi Ai

Built ML pipelines for risk scoring, fraud detection, and user behavior modeling in financial transaction and customer analytics use cases. Improved fraud detection performance by increasing F1-score by 15% through feature engineering, model validation, and iterative experimentation. Deployed time-series forecasting models (SARIMA, Prophet) for transaction-volume and liquidity prediction, reducing MAPE by 20%. Designed feature engineering workflows for financial signals (rolling statistics, volatility, seasonality, anomaly detection), decreasing false positives in fraud triggers..

NLP Engineer

Kontur

Built and deployed NLP services for user query classification, intent recognition, and sentiment analysis in customer-facing workflows. Automated key data extraction from contracts and business documents using NLP and LLM-based pipelines, reducing processing time by 10-15%. Fine-tuned and quantized LLMs with PEFT and LoRA for efficient deployment in resource-constrained environments. Designed hybrid learning workflows leveraging LLM-assisted data labeling and BERT-family retraining, reducing annotation overhead by 30%..

CV Engineer

VisionLabs

Designed and deployed a CT-based abdominal organ and pathology segmentation system for medical imaging workflows processing DICOM data. Developed preprocessing, inference, and post-processing pipelines for volumetric CT studies, supporting reproducible model evaluation on real-world clinical data. Improved inference latency by 20% while achieving Dice score 0.80 and ROC AUC 0.98 on clinical validation datasets. Integrated Kafka-based DICOM streaming into diagnostic reporting pipeline, enabling asynchronous message processing between imaging systems..

Senior ML Engineer

Skoltech (Skolkovo Institute of Science and Technology)— Artificial Intelligence Center

Architected and built AgentCreator, an end-to-end fullstack AI-agent platform for chat-based LLM assistance, autonomous agentic workflows, tools, skills, artifacts, logs, benchmarks, and Dockerized deployment. Designed the core system architecture across FastAPI backend, async SQLAlchemy/Postgres storage, LLM tool-calling runtime, and Docker Compose infrastructure. Implemented a custom agentic runtime with a frontier assistant, executor workflow, autonomous tool-calling loop, guardrails, tool-result history management, and OpenAI-compatible function calling without relying on LangChain..

Senior Data Scientist

BP

Architected and deployed a production-grade information extraction pipeline combining Azure Document Intelligence OCR/layout parsing, Azure Content Understanding, and Azure OpenAI to automate structured metadata extraction from complex business documents, achieving 95%+ extraction accuracy. Engineered scalable backend business logic around a robust domain model to handle automated data ingestion and YAML-driven extraction configurations with strict schema validation for downstream systems. Developed and trained gradient-boosted regression models to predict critical oil properties, succes

BCs. Applied Mathematics and Computer Science

Saint-Petersburg State University

MSc. Fundamental Informatics and Information Technologies

Saint-Petersburg State University