AI/ML Team Lead · Multi-Agent Systems · MCP · RAG · Enterprise AI

I am a forward deployed AI engineer with experience in building multi-agent systems and enterprise AI solutions. Currently working as AI/ML Team Lead at fintech company. Before that I spent 6 years at Gazprom Neft as Head of Data Quality, where I was building data platforms and leading teams.

I specialize in MCP protocol, RAG systems, and Google ADK framework. I like to work on problems where AI agents need to interact with real systems — databases, APIs, and business tools.

┌─────────────────────────────────────────────────────┐
│                   Orchestration                      │
│        Google ADK · LangChain · LangGraph            │
├─────────────────────────────────────────────────────┤
│                     MCP Layer                        │
│     Model Context Protocol · Custom Servers           │
├─────────────────────────────────────────────────────┤
│                     RAG Layer                        │
│    LlamaIndex · Vector Search · Embeddings            │
├─────────────────────────────────────────────────────┤
│                  Deep Learning                        │
│           PyTorch · ONNX · GPU Opt                   │
├─────────────────────────────────────────────────────┤
│                    Platform                          │
│      Python · Docker · FastAPI · PostgreSQL           │
└─────────────────────────────────────────────────────┘

Languages: Python, SQL, Bash Frameworks: PyTorch, FastAPI, LlamaIndex, LangChain, Google ADK Infra: Docker, AWS, S3, PostgreSQL, SQLite Data: Pandas, Spark, DBT

• mcp-server-kbsearch — MCP server for semantic search. Using LlamaIndex embeddings and auto-ingest from S3. You can search through knowledge base with natural language queries.

• mcp-server-dbsearch — MCP server for database search. Auto-loads Excel and CSV files from S3 to SQLite, provides REST API for querying. Good for when you need to search through structured data.

• dbhub_example — Financial analyst agent built with Google ADK and MCP. It can query PostgreSQL database and verify SQL execution. Shows how to build agent that works with real financial data.

• asch-chat-bot — Multi-agent chat bot with product selection. Uses MCP tools, PostgreSQL with Alembic migrations, deployed in Docker. Each agent handles different part of conversation.

• ya-assistant-api — Integration with Yandex AI Assistant API. Prototyping different tools and workflows for Yandex platform. Shows how to extend LLM capabilities with custom tools.

• moleskin — PyTorch training pipelines for melanoma classification on ISIC 2019 dataset. Includes GPU inference optimization and experiment tracking.

• mediag — Deep learning research for medical image analysis. Working on cancer diagnosis using cell classification models.

Recent work includes:

• Building multi-agent systems with MCP protocol for enterprise use
• Designing RAG pipelines with vector search and document retrieval
• Leading ML team of 5+ engineers
• Implementing data quality frameworks for petabyte-scale systems

• CDMP — Certified Data Management Professional
• IBM Deep Learning Specialization — Neural Networks
• MBA — Master of Business Administration
• Systems Analysis — Bachelor degree

Email: itai.kogtev@gmail.com LinkedIn: linkedin.com/in/vkogtev Location: Israel (open to remote) Available for freelance projects