AI Systems Engineer | Data Scientist | ML | Multimodal AI

AI Systems Engineer and Data Scientist specializing in designing and building end-to-end AI systems, scalable architectures, and production-ready platforms.

Experience spans the full machine learning lifecycle: data processing → dataset engineering → model training → evaluation → deployment → system integration.

Focus is on system-level AI, where multiple models, pipelines, and tools are orchestrated into unified intelligent systems rather than used in isolation.

Strong emphasis on architecture, orchestration, and integration of multimodal AI systems.

• Agent-based AI systems
• Multimodal AI (computer vision, NLP, audio, video)
• Dynamic routing and orchestration
• AI system architecture and pipeline design
• End-to-end machine learning systems
• Explainable and controllable AI systems
• Model optimization and experimental validation

Advanced agent-based AI platform for multimodal processing, dynamic routing, and system-level orchestration.

System architecture:

User → Murzik (Agent) → Quantum Orchestrator → Optimization Layer → Routing Layer → Tools/Models → Post-processing → Explanation → Output

• Murzik — multi-layer AI agent (interaction, execution, explanation)
• Dynamic routing system (context-aware pipeline selection)
• Quantum-inspired orchestration (multi-path reasoning and reranking)
• Multimodal pipelines (vision, text, audio, video)
• AI Console (YOLO, LLM, ASR, CV pipelines, experimental modes)
• Logging and explainability system

Murzik 2 (Quantum Mode)

• Dual-model architecture (LLM + multimodal model)
• Embedding-level transformations
• Alternative reasoning pipeline
• Research-oriented experimental framework

• System-level AI architecture
• Integration of multiple models into unified systems
• Adaptive execution and real-time decision logic
• Production-ready systems with research direction

End-to-end machine learning system for analyzing exoplanet candidates using NASA data.

• Data collection via NASA Exoplanet Archive API
• Feature engineering and preprocessing pipelines
• Handling imbalanced datasets (SMOTE)
• Model training and optimization (Random Forest, GridSearchCV)
• Evaluation using precision, recall, F1-score
• Experimental validation and prediction generation

• Participation in global NASA hackathon
• Solving real-world Earth and space-related challenges
• Applied data-driven and system-level problem-solving approaches
• Work under time constraints in a global environment
• Recognition: Galactic Problem Solver

• Full-cycle AI systems (data → models → deployment)
• Computer vision systems and ML pipelines
• Custom neural architectures and optimization
• Production-level system design and integration

Python | PyTorch | TensorFlow | Scikit-learn
Computer Vision | NLP | LLM

FastAPI | Docker | SQL

Multimodal AI | ML Pipelines | System Architecture

Email: srumyantseva7@gmail.com
LinkedIn: https://www.linkedin.com/in/svetlana-rumyantseva-5b41962b9
Kaggle: https://www.kaggle.com/svetlanarumyantseva7
Telegram: https://t.me/Svetlana_KostraTana