Open to opportunities

Svetlana Rumyantseva

@svetlanarumyantseva

AI Systems Engineer building production-ready LLM, computer vision, and multimodal platforms with agent-based architectures.

Panama

Message

What I'm looking for

I’m looking for a team where I can build production-grade LLM + multimodal systems with agent-based architectures, dynamic routing, and rigorous evaluation. I want ownership of end-to-end lifecycle work and structured experimentation to improve reliability and performance.

I’m an AI Systems Engineer and Data Scientist focused on designing and developing end-to-end AI systems. I specialize in LLM-based architectures, computer vision, and multimodal pipelines, with a strong emphasis on system-level reliability and performance.

In my current work on Golden Dragon AI, I’m building an advanced multimodal, adaptive AI platform with next-generation architecture. I develop agent-based systems (e.g., Murzik) for coordination across multiple components, along with dynamic routing and orchestration so the system can select and execute pipelines in a context-aware way.

Across projects, I work as a sole AI engineer and own the full lifecycle—from system design and data pipelines to model training, evaluation, deployment, and continuous improvement. I’m especially driven by evaluation and benchmarking: structured experimentation to improve reasoning quality, model reliability, and reduce latency (by ~20–30%) through batching and efficient processing.

I also build production-ready applications and tools using frameworks like Streamlit and React, and I’ve delivered domain-focused systems such as a Veterinary AI System for horse anatomy analysis and biometrical movement understanding. My engineering philosophy is simple: rigorous experiments, robust architectures, and practical deployment over “demo-only” results.

Experience

Work history, roles, and key accomplishments

Current

AI Systems Engineer

Current

Golden Dragon AI

Mar 2026 - Present (3 months)

Designed and developed a scalable, production-grade AI platform combining LLMs, computer vision, and multimodal pipelines using an agent-based architecture with dynamic routing and orchestration. Built the Murzik agent for system-level coordination and reduced inference/pipeline latency by ~20–30% through batching and performance optimization.

LLM Systems RAG Agent Architectures Prompt Engineering Computer Vision Performance Optimization System Deployment

Education

Degrees, certifications, and relevant coursework

University of Artificial Intelligence

Diploma in Data Science, Data Science

2020 - 2022

Activities and societies: Core areas included Machine Learning & Deep Learning, Neural Networks (CNN/RNN/Autoencoders/GANs), Computer Vision (YOLO/segmentation/tracking), NLP (seq2seq/attention/transformers), Reinforcement Learning (Q-learning/policy gradients), and Time Series & Audio Processing.

Completed a comprehensive Diploma in Data Science covering model training, evaluation, and applied AI. Studied machine learning, deep learning, computer vision, NLP, reinforcement learning, and time series/audio processing.